{"fetched_at":"2026-04-23T08:26:30.195Z","cache_expires_at":"2026-04-23T09:26:30.195Z","count":500,"model_count":134,"variants":[{"variant_id":"openrouter::openrouter/auto::tools+reasoning","model_id":"openrouter/auto","canonical_id":"openrouter/auto","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Auto Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":2000000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":-1000000,"pricing_output_per_1m":-1000000,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","audio","file","video"],"output_modalities":["text","image"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1141,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.689027+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::openrouter/auto::structured","model_id":"openrouter/auto","canonical_id":"openrouter/auto","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Auto Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":2000000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":-1000000,"pricing_output_per_1m":-1000000,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","audio","file","video"],"output_modalities":["text","image"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":610,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.698156+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::openrouter/auto::tools","model_id":"openrouter/auto","canonical_id":"openrouter/auto","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Auto Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":2000000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":-1000000,"pricing_output_per_1m":-1000000,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","audio","file","video"],"output_modalities":["text","image"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T11:14:10.003880+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::openrouter/auto::base","model_id":"openrouter/auto","canonical_id":"openrouter/auto","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Auto Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":2000000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":-1000000,"pricing_output_per_1m":-1000000,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","audio","file","video"],"output_modalities":["text","image"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:52:09.986755+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/auto::reasoning","model_id":"openrouter/auto","canonical_id":"openrouter/auto","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Auto Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":2000000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":-1000000,"pricing_output_per_1m":-1000000,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","audio","file","video"],"output_modalities":["text","image"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":7000,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.911526+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::openrouter/bodybuilder::base","model_id":"openrouter/bodybuilder","canonical_id":"openrouter/bodybuilder","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Body Builder (beta)","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":-1000000,"pricing_output_per_1m":-1000000,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":885,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.219818+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/auto::low_temp","model_id":"openrouter/auto","canonical_id":"openrouter/auto","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Auto Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":2000000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":-1000000,"pricing_output_per_1m":-1000000,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","audio","file","video"],"output_modalities":["text","image"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T11:14:13.439544+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_completion_tokens","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p","web_search_options"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/bodybuilder::low_temp","model_id":"openrouter/bodybuilder","canonical_id":"openrouter/bodybuilder","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Body Builder (beta)","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":-1000000,"pricing_output_per_1m":-1000000,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":777,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.982260+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-20b::tools+reasoning","model_id":"openai/gpt-oss-20b:free","canonical_id":"openai/gpt-oss-20b","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-20b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":380,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.395462+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-3-llama-3.1-405b::search","model_id":"nousresearch/hermes-3-llama-3.1-405b:free","canonical_id":"nousresearch/hermes-3-llama-3.1-405b","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nous: Hermes 3 405B Instruct (free)","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":405,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":242,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.843605+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-480b-a35b-07-25::search","model_id":"qwen/qwen3-coder:free","canonical_id":"qwen/qwen3-coder-480b-a35b-07-25","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder 480B A35B (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262000,"max_completion_tokens":262000,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":480,"active_param_count_b":35,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":1657,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.896250+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-12b-v2-vl::tools","model_id":"nvidia/nemotron-nano-12b-v2-vl:free","canonical_id":"nvidia/nemotron-nano-12b-v2-vl","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 12B 2 VL (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":128000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":false,"unavailable_reason":"missing_capabilities:tools","last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-03-15T19:48:48.939000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:22.755055+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.2-3b-instruct::low_temp","model_id":"meta-llama/llama-3.2-3b-instruct:free","canonical_id":"meta-llama/llama-3.2-3b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.2 3B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3.2","decoder_type":"Dense","attention":"GQA","arch_highlight":"Reference small-model Llama architecture with tied embeddings.","tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:48.001422+00:00","status_source":"verify","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-next-80b-a3b-instruct-2509::base","model_id":"qwen/qwen3-next-80b-a3b-instruct:free","canonical_id":"qwen/qwen3-next-80b-a3b-instruct-2509","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Next 80B A3B Instruct (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":false,"unavailable_reason":"model_not_found_in_or","last_checked_at":"2026-04-12T01:22:53.086255+00:00","last_latency_ms":null,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:06:01.578717+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.3-70b-instruct::low_temp","model_id":"meta-llama/llama-3.3-70b-instruct:free","canonical_id":"meta-llama/llama-3.3-70b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.3 70B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":65536,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.983870+00:00","status_source":"verify","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-480b-a35b-07-25::base","model_id":"qwen/qwen3-coder:free","canonical_id":"qwen/qwen3-coder-480b-a35b-07-25","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder 480B A35B (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262000,"max_completion_tokens":262000,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":480,"active_param_count_b":35,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":false,"unavailable_reason":"model_not_found_in_or","last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:57.467080+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-4b-it::low_temp","model_id":"google/gemma-3-4b-it:free","canonical_id":"google/gemma-3-4b-it","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.873459+00:00","status_source":"verify","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-nano-30b-a3b::low_temp","model_id":"nvidia/nemotron-3-nano-30b-a3b:free","canonical_id":"nvidia/nemotron-3-nano-30b-a3b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Nano 30B A3B (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":256000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":242,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.909945+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-super-120b-a12b-20230311::structured","model_id":"nvidia/nemotron-3-super-120b-a12b:free","canonical_id":"nvidia/nemotron-3-super-120b-a12b-20230311","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Super (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1085,"first_unavailable_at":"2026-03-15T23:45:52.769000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.790652+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::openrouter/free::reasoning","model_id":"openrouter/free","canonical_id":"openrouter/free","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Free Models Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":200000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2429,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.424702+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-nano-30b-a3b::search","model_id":"nvidia/nemotron-3-nano-30b-a3b:free","canonical_id":"nvidia/nemotron-3-nano-30b-a3b","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Nano 30B A3B (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":256000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":258,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.845038+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-20b::base","model_id":"openai/gpt-oss-20b:free","canonical_id":"openai/gpt-oss-20b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-20b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:04:29.332455+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-120b::reasoning","model_id":"openai/gpt-oss-120b:free","canonical_id":"openai/gpt-oss-120b","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-120b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":967,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.714543+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::minimax/minimax-m2.5-20260211::search+tools","model_id":"minimax/minimax-m2.5:free","canonical_id":"minimax/minimax-m2.5-20260211","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax M2.5 (free)","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":196608,"max_completion_tokens":8192,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":7113,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.930006+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-super-120b-a12b-20230311::search+tools","model_id":"nvidia/nemotron-3-super-120b-a12b:free","canonical_id":"nvidia/nemotron-3-super-120b-a12b-20230311","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Super (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":330,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.934086+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-m2.5-20260211::reasoning","model_id":"minimax/minimax-m2.5:free","canonical_id":"minimax/minimax-m2.5-20260211","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax M2.5 (free)","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":196608,"max_completion_tokens":8192,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1893,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.417815+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-9b-v2::search+tools","model_id":"nvidia/nemotron-nano-9b-v2:free","canonical_id":"nvidia/nemotron-nano-9b-v2","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 9B V2 (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":175,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.936223+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/free::low_temp","model_id":"openrouter/free","canonical_id":"openrouter/free","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Free Models Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":200000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1213,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.883143+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-m2.5-20260211::search","model_id":"minimax/minimax-m2.5:free","canonical_id":"minimax/minimax-m2.5-20260211","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax M2.5 (free)","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":196608,"max_completion_tokens":8192,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":337,"first_unavailable_at":"2026-04-13T20:35:56.784251+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.841649+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-9b-v2::reasoning","model_id":"nvidia/nemotron-nano-9b-v2:free","canonical_id":"nvidia/nemotron-nano-9b-v2","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 9B V2 (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":843,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.694815+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-120b::search+tools","model_id":"openai/gpt-oss-120b:free","canonical_id":"openai/gpt-oss-120b","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-120b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":3389,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.937296+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.3-70b-instruct::search","model_id":"meta-llama/llama-3.3-70b-instruct:free","canonical_id":"meta-llama/llama-3.3-70b-instruct","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.3 70B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":65536,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":346,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.840052+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-next-80b-a3b-instruct-2509::search","model_id":"qwen/qwen3-next-80b-a3b-instruct:free","canonical_id":"qwen/qwen3-next-80b-a3b-instruct-2509","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Next 80B A3B Instruct (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":287,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.899796+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.5-air::tools+reasoning","model_id":"z-ai/glm-4.5-air:free","canonical_id":"z-ai/glm-4.5-air","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.5 Air (free)","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":96000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1137,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.403193+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.2-3b-instruct::search","model_id":"meta-llama/llama-3.2-3b-instruct:free","canonical_id":"meta-llama/llama-3.2-3b-instruct","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.2 3B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":367,"first_unavailable_at":null,"arch_id":"llama-3.2","decoder_type":"Dense","attention":"GQA","arch_highlight":"Reference small-model Llama architecture with tied embeddings.","tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.838511+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.5-air::base","model_id":"z-ai/glm-4.5-air:free","canonical_id":"z-ai/glm-4.5-air","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.5 Air (free)","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":96000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:06:00.657336+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-12b-it::low_temp","model_id":"google/gemma-3-12b-it:free","canonical_id":"google/gemma-3-12b-it","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 12B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.874725+00:00","status_source":"verify","claimed_capabilities":["max_tokens","seed","stop","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-120b::low_temp","model_id":"openai/gpt-oss-120b:free","canonical_id":"openai/gpt-oss-120b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-120b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.723910+00:00","status_source":"verify","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-nano-30b-a3b::search+tools","model_id":"nvidia/nemotron-3-nano-30b-a3b:free","canonical_id":"nvidia/nemotron-3-nano-30b-a3b","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Nano 30B A3B (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":256000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":470,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.932988+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3n-e4b-it::base","model_id":"google/gemma-3n-e4b-it:free","canonical_id":"google/gemma-3n-e4b-it","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3n 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":8192,"max_completion_tokens":2048,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:04:27.833015+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/free::tools","model_id":"openrouter/free","canonical_id":"openrouter/free","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Free Models Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":200000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:28.303651+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-12b-v2-vl::base","model_id":"nvidia/nemotron-nano-12b-v2-vl:free","canonical_id":"nvidia/nemotron-nano-12b-v2-vl","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 12B 2 VL (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":128000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:55.509474+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-m2.5-20260211::tools+reasoning","model_id":"minimax/minimax-m2.5:free","canonical_id":"minimax/minimax-m2.5-20260211","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax M2.5 (free)","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":196608,"max_completion_tokens":8192,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":4972,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.002893+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-12b-v2-vl::search+tools","model_id":"nvidia/nemotron-nano-12b-v2-vl:free","canonical_id":"nvidia/nemotron-nano-12b-v2-vl","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 12B 2 VL (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":128000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":182,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.935171+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/elephant-alpha::search","model_id":"openrouter/elephant-alpha","canonical_id":"openrouter/elephant-alpha","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Elephant","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-13T19:43:02.792642+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":601,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.892067+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-next-80b-a3b-instruct-2509::search+tools","model_id":"qwen/qwen3-next-80b-a3b-instruct:free","canonical_id":"qwen/qwen3-next-80b-a3b-instruct-2509","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Next 80B A3B Instruct (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":419,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.942381+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-3-llama-3.1-405b::low_temp","model_id":"nousresearch/hermes-3-llama-3.1-405b:free","canonical_id":"nousresearch/hermes-3-llama-3.1-405b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nous: Hermes 3 405B Instruct (free)","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":405,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:48.077591+00:00","status_source":"verify","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-3-llama-3.1-405b::base","model_id":"nousresearch/hermes-3-llama-3.1-405b:free","canonical_id":"nousresearch/hermes-3-llama-3.1-405b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nous: Hermes 3 405B Instruct (free)","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":405,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:52.116957+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.5-air::search","model_id":"z-ai/glm-4.5-air:free","canonical_id":"z-ai/glm-4.5-air","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.5 Air (free)","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":96000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":715,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.903403+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-super-120b-a12b-20230311::low_temp","model_id":"nvidia/nemotron-3-super-120b-a12b:free","canonical_id":"nvidia/nemotron-3-super-120b-a12b-20230311","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Super (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":233,"first_unavailable_at":"2026-03-15T23:45:52.769000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.785758+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-12b-it::tools","model_id":"google/gemma-3-12b-it:free","canonical_id":"google/gemma-3-12b-it","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 12B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":null,"unavailable_reason":null,"last_checked_at":null,"last_latency_ms":null,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:20.566785+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","seed","stop","temperature","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-12b-it::base","model_id":"google/gemma-3-12b-it:free","canonical_id":"google/gemma-3-12b-it","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 12B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:52.079323+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","seed","stop","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-20b::reasoning","model_id":"openai/gpt-oss-20b:free","canonical_id":"openai/gpt-oss-20b","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-20b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":459,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.781481+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::google/gemma-3-27b-it::base","model_id":"google/gemma-3-27b-it:free","canonical_id":"google/gemma-3-27b-it","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 27B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:55.058162+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3n-e4b-it::search+tools","model_id":"google/gemma-3n-e4b-it:free","canonical_id":"google/gemma-3n-e4b-it","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3n 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":8192,"max_completion_tokens":2048,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":1112,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.918369+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","temperature","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-120b::tools+reasoning","model_id":"openai/gpt-oss-120b:free","canonical_id":"openai/gpt-oss-120b","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-120b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":434,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.392894+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-super-120b-a12b-20230311::reasoning","model_id":"nvidia/nemotron-3-super-120b-a12b:free","canonical_id":"nvidia/nemotron-3-super-120b-a12b-20230311","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Super (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":406,"first_unavailable_at":"2026-03-15T23:45:52.769000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.383016+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::google/gemma-3n-e4b-it::search","model_id":"google/gemma-3n-e4b-it:free","canonical_id":"google/gemma-3n-e4b-it","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3n 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":8192,"max_completion_tokens":2048,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":4168,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.827703+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","temperature","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-9b-v2::tools","model_id":"nvidia/nemotron-nano-9b-v2:free","canonical_id":"nvidia/nemotron-nano-9b-v2","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 9B V2 (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:22.371603+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-nano-30b-a3b::reasoning","model_id":"nvidia/nemotron-3-nano-30b-a3b:free","canonical_id":"nvidia/nemotron-3-nano-30b-a3b","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Nano 30B A3B (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":256000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1164,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.502231+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-next-80b-a3b-instruct-2509::tools","model_id":"qwen/qwen3-next-80b-a3b-instruct:free","canonical_id":"qwen/qwen3-next-80b-a3b-instruct-2509","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Next 80B A3B Instruct (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":false,"unavailable_reason":"model_not_found_in_or","last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-04-12T01:23:08.011605+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:17:47.605867+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-4b-it::base","model_id":"google/gemma-3-4b-it:free","canonical_id":"google/gemma-3-4b-it","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:54.274801+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.5-air::tools","model_id":"z-ai/glm-4.5-air:free","canonical_id":"z-ai/glm-4.5-air","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.5 Air (free)","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":96000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:26.182407+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::google/gemma-3-12b-it::search+tools","model_id":"google/gemma-3-12b-it:free","canonical_id":"google/gemma-3-12b-it","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 12B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":450,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.913086+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","seed","stop","temperature","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-27b-it::search","model_id":"google/gemma-3-27b-it:free","canonical_id":"google/gemma-3-27b-it","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 27B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":744,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.823145+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3n-e4b-it::low_temp","model_id":"google/gemma-3n-e4b-it:free","canonical_id":"google/gemma-3n-e4b-it","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3n 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":8192,"max_completion_tokens":2048,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.785909+00:00","status_source":"verify","claimed_capabilities":["max_tokens","response_format","seed","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-20b::search","model_id":"openai/gpt-oss-20b:free","canonical_id":"openai/gpt-oss-20b","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-20b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":2835,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.890421+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/elephant-alpha::base","model_id":"openrouter/elephant-alpha","canonical_id":"openrouter/elephant-alpha","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Elephant","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-13T19:43:02.792642+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":null,"unavailable_reason":null,"last_checked_at":null,"last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:17:17.857547+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-12b-v2-vl::low_temp","model_id":"nvidia/nemotron-nano-12b-v2-vl:free","canonical_id":"nvidia/nemotron-nano-12b-v2-vl","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 12B 2 VL (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":128000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.483182+00:00","status_source":"verify","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/free::base","model_id":"openrouter/free","canonical_id":"openrouter/free","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Free Models Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":200000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1323,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:06:24.458575+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/elephant-alpha::search+tools","model_id":"openrouter/elephant-alpha","canonical_id":"openrouter/elephant-alpha","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Elephant","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-13T19:43:02.792642+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":682,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.939420+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-27b-it::search+tools","model_id":"google/gemma-3-27b-it:free","canonical_id":"google/gemma-3-27b-it","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 27B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":640,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.914306+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/free::tools+reasoning","model_id":"openrouter/free","canonical_id":"openrouter/free","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Free Models Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":200000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":2173,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.012972+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-9b-v2::base","model_id":"nvidia/nemotron-nano-9b-v2:free","canonical_id":"nvidia/nemotron-nano-9b-v2","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 9B V2 (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:54.972383+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/elephant-alpha::tools","model_id":"openrouter/elephant-alpha","canonical_id":"openrouter/elephant-alpha","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Elephant","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-13T19:43:02.792642+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":null,"unavailable_reason":null,"last_checked_at":null,"last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:17:19.635626+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.5-air::low_temp","model_id":"z-ai/glm-4.5-air:free","canonical_id":"z-ai/glm-4.5-air","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.5 Air (free)","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":96000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.734270+00:00","status_source":"verify","claimed_capabilities":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-12b-it::search","model_id":"google/gemma-3-12b-it:free","canonical_id":"google/gemma-3-12b-it","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 12B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":348,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.820953+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","seed","stop","temperature","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-4b-it::search","model_id":"google/gemma-3-4b-it:free","canonical_id":"google/gemma-3-4b-it","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":942,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.824565+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-9b-v2::structured","model_id":"nvidia/nemotron-nano-9b-v2:free","canonical_id":"nvidia/nemotron-nano-9b-v2","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 9B V2 (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":233,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.195815+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemma-3-4b-it::search+tools","model_id":"google/gemma-3-4b-it:free","canonical_id":"google/gemma-3-4b-it","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":798,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.915523+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-27b-it::tools","model_id":"google/gemma-3-27b-it:free","canonical_id":"google/gemma-3-27b-it","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 27B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":false,"unavailable_reason":"missing_capabilities:tools","last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-03-15T19:50:01.835000+00:00","arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:20.480510+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-12b-v2-vl::search","model_id":"nvidia/nemotron-nano-12b-v2-vl:free","canonical_id":"nvidia/nemotron-nano-12b-v2-vl","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 12B 2 VL (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":128000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":223,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.847676+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-3-llama-3.1-405b::tools","model_id":"nousresearch/hermes-3-llama-3.1-405b:free","canonical_id":"nousresearch/hermes-3-llama-3.1-405b","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nous: Hermes 3 405B Instruct (free)","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":405,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":null,"unavailable_reason":null,"last_checked_at":null,"last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:22.054931+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-20b::search+tools","model_id":"openai/gpt-oss-20b:free","canonical_id":"openai/gpt-oss-20b","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-20b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":569,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.938361+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/free::structured","model_id":"openrouter/free","canonical_id":"openrouter/free","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Free Models Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":200000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":7604,"first_unavailable_at":"2026-04-12T13:14:11.279531+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.881683+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemma-4-31b-it-20260402::search","model_id":"google/gemma-4-31b-it:free","canonical_id":"google/gemma-4-31b-it-20260402","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 31B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Gemma","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":31,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-08T01:41:18.950874+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":415,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.830938+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-m2.5-20260211::tools","model_id":"minimax/minimax-m2.5:free","canonical_id":"minimax/minimax-m2.5-20260211","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax M2.5 (free)","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":196608,"max_completion_tokens":8192,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":false,"unavailable_reason":"model_not_found_in_or","last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-04-12T01:23:08.011605+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:20.327168+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/free::search+tools","model_id":"openrouter/free","canonical_id":"openrouter/free","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Free Models Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":200000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":565,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.940399+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.3-70b-instruct::base","model_id":"meta-llama/llama-3.3-70b-instruct:free","canonical_id":"meta-llama/llama-3.3-70b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.3 70B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":65536,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:52.333708+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-super-120b-a12b-20230311::base","model_id":"nvidia/nemotron-3-super-120b-a12b:free","canonical_id":"nvidia/nemotron-3-super-120b-a12b-20230311","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Super (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":240,"first_unavailable_at":"2026-03-15T23:45:52.769000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:06:03.419223+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-4-31b-it-20260402::search+tools","model_id":"google/gemma-4-31b-it:free","canonical_id":"google/gemma-4-31b-it-20260402","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 31B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Gemma","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":31,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-08T01:41:18.950874+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":709,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.921008+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-super-120b-a12b-20230311::tools","model_id":"nvidia/nemotron-3-super-120b-a12b:free","canonical_id":"nvidia/nemotron-3-super-120b-a12b-20230311","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Super (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":false,"unavailable_reason":"model_not_found_in_or","last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-03-15T23:45:52.769000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:24.861129+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.2-3b-instruct::tools","model_id":"meta-llama/llama-3.2-3b-instruct:free","canonical_id":"meta-llama/llama-3.2-3b-instruct","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.2 3B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":null,"unavailable_reason":null,"last_checked_at":null,"last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3.2","decoder_type":"Dense","attention":"GQA","arch_highlight":"Reference small-model Llama architecture with tied embeddings.","tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:20.007955+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openrouter/free::search","model_id":"openrouter/free","canonical_id":"openrouter/free","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Free Models Router","org":"openrouter","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":200000,"max_completion_tokens":null,"tokenizer":"Router","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":1560,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.893601+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-120b::tools","model_id":"openai/gpt-oss-120b:free","canonical_id":"openai/gpt-oss-120b","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-120b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:24.249500+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.2-3b-instruct::search+tools","model_id":"meta-llama/llama-3.2-3b-instruct:free","canonical_id":"meta-llama/llama-3.2-3b-instruct","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.2 3B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":353,"first_unavailable_at":null,"arch_id":"llama-3.2","decoder_type":"Dense","attention":"GQA","arch_highlight":"Reference small-model Llama architecture with tied embeddings.","tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.927284+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-120b::search","model_id":"openai/gpt-oss-120b:free","canonical_id":"openai/gpt-oss-120b","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-120b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":516,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.888962+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-9b-v2::search","model_id":"nvidia/nemotron-nano-9b-v2:free","canonical_id":"nvidia/nemotron-nano-9b-v2","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 9B V2 (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":395,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.849160+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-nano-30b-a3b::tools+reasoning","model_id":"nvidia/nemotron-3-nano-30b-a3b:free","canonical_id":"nvidia/nemotron-3-nano-30b-a3b","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Nano 30B A3B (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":256000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1093,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.108892+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.5-air::reasoning","model_id":"z-ai/glm-4.5-air:free","canonical_id":"z-ai/glm-4.5-air","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.5 Air (free)","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":96000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1845,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.786528+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-next-80b-a3b-instruct-2509::structured","model_id":"qwen/qwen3-next-80b-a3b-instruct:free","canonical_id":"qwen/qwen3-next-80b-a3b-instruct-2509","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Next 80B A3B Instruct (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":435,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.185989+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-super-120b-a12b-20230311::search","model_id":"nvidia/nemotron-3-super-120b-a12b:free","canonical_id":"nvidia/nemotron-3-super-120b-a12b-20230311","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Super (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":371,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.846323+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.5-air::search+tools","model_id":"z-ai/glm-4.5-air:free","canonical_id":"z-ai/glm-4.5-air","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.5 Air (free)","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":96000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":8651,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.978324+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-m2.5-20260211::base","model_id":"minimax/minimax-m2.5:free","canonical_id":"minimax/minimax-m2.5-20260211","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax M2.5 (free)","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":196608,"max_completion_tokens":8192,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1554,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:59.528298+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-4-26b-a4b-it-20260403::search+tools","model_id":"google/gemma-4-26b-a4b-it:free","canonical_id":"google/gemma-4-26b-a4b-it-20260403","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 26B A4B  (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Gemma","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":26,"active_param_count_b":4,"is_moe":true,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-08T01:41:18.950874+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":420,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.919942+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.2-3b-instruct::base","model_id":"meta-llama/llama-3.2-3b-instruct:free","canonical_id":"meta-llama/llama-3.2-3b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.2 3B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3.2","decoder_type":"Dense","attention":"GQA","arch_highlight":"Reference small-model Llama architecture with tied embeddings.","tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-15T02:33:14.111708+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-9b-v2::low_temp","model_id":"nvidia/nemotron-nano-9b-v2:free","canonical_id":"nvidia/nemotron-nano-9b-v2","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 9B V2 (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.792794+00:00","status_source":"verify","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-480b-a35b-07-25::tools","model_id":"qwen/qwen3-coder:free","canonical_id":"qwen/qwen3-coder-480b-a35b-07-25","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder 480B A35B (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262000,"max_completion_tokens":262000,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":480,"active_param_count_b":35,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":false,"unavailable_reason":"model_not_found_in_or","last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-04-12T01:23:08.011605+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:23.649387+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-120b::base","model_id":"openai/gpt-oss-120b:free","canonical_id":"openai/gpt-oss-120b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-120b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:06:23.166185+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.3-70b-instruct::tools","model_id":"meta-llama/llama-3.3-70b-instruct:free","canonical_id":"meta-llama/llama-3.3-70b-instruct","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.3 70B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":65536,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:21.157886+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::google/gemma-3-27b-it::low_temp","model_id":"google/gemma-3-27b-it:free","canonical_id":"google/gemma-3-27b-it","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 27B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.880432+00:00","status_source":"verify","claimed_capabilities":["max_tokens","response_format","seed","stop","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-9b-v2::tools+reasoning","model_id":"nvidia/nemotron-nano-9b-v2:free","canonical_id":"nvidia/nemotron-nano-9b-v2","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 9B V2 (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":907,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.308002+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-nano-30b-a3b::base","model_id":"nvidia/nemotron-3-nano-30b-a3b:free","canonical_id":"nvidia/nemotron-3-nano-30b-a3b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Nano 30B A3B (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":256000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":181,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:56.863375+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-480b-a35b-07-25::search+tools","model_id":"qwen/qwen3-coder:free","canonical_id":"qwen/qwen3-coder-480b-a35b-07-25","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder 480B A35B (free)","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262000,"max_completion_tokens":262000,"tokenizer":"Qwen3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":480,"active_param_count_b":35,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":1126,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.941379+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-m2.5-20260211::low_temp","model_id":"minimax/minimax-m2.5:free","canonical_id":"minimax/minimax-m2.5-20260211","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax M2.5 (free)","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":196608,"max_completion_tokens":8192,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1795,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.816581+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","temperature","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-3-nano-30b-a3b::tools","model_id":"nvidia/nemotron-3-nano-30b-a3b:free","canonical_id":"nvidia/nemotron-3-nano-30b-a3b","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron 3 Nano 30B A3B (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":256000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:21.871318+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.3-70b-instruct::search+tools","model_id":"meta-llama/llama-3.3-70b-instruct:free","canonical_id":"meta-llama/llama-3.3-70b-instruct","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.3 70B Instruct (free)","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":65536,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":388,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.928503+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/nemotron-nano-12b-v2-vl::reasoning","model_id":"nvidia/nemotron-nano-12b-v2-vl:free","canonical_id":"nvidia/nemotron-nano-12b-v2-vl","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Nemotron Nano 12B 2 VL (free)","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":128000,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":681,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.604478+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-20b::low_temp","model_id":"openai/gpt-oss-20b:free","canonical_id":"openai/gpt-oss-20b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-20b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:29:47.725078+00:00","status_source":"verify","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-4-26b-a4b-it-20260403::search","model_id":"google/gemma-4-26b-a4b-it:free","canonical_id":"google/gemma-4-26b-a4b-it-20260403","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 26B A4B  (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Gemma","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":26,"active_param_count_b":4,"is_moe":true,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-08T01:41:18.950874+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:37:55.608791+00:00","last_latency_ms":615,"first_unavailable_at":"2026-04-13T20:35:56.784251+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:59.829593+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-20b::tools","model_id":"openai/gpt-oss-20b:free","canonical_id":"openai/gpt-oss-20b","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: gpt-oss-20b (free)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":8192,"tokenizer":"GPT","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:23.370814+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","seed","stop","temperature","tool_choice","tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-3-llama-3.1-405b::search+tools","model_id":"nousresearch/hermes-3-llama-3.1-405b:free","canonical_id":"nousresearch/hermes-3-llama-3.1-405b","config_key":"search+tools","variant_key":"search_tools","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search + Tools)","variant_note":"Flags: search+tools","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nous: Hermes 3 405B Instruct (free)","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":405,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-14T00:31:31.194404+00:00","last_latency_ms":267,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-14T00:31:40.931532+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","stop","temperature","top_k","top_p"],"required_capabilities":["tools","web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3n-e4b-it::tools","model_id":"google/gemma-3n-e4b-it:free","canonical_id":"google/gemma-3n-e4b-it","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3n 4B (free)","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":8192,"max_completion_tokens":2048,"tokenizer":"Other","pricing_input_per_1m":0,"pricing_output_per_1m":0,"tags":"free text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","temperature","top_p"],"rate_limit_rpm":20,"rate_limit_rpd":200,"rate_limit_tpm":null,"rate_limit_source":"tier_derived","provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":null,"unavailable_reason":null,"last_checked_at":null,"last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:22.314190+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","temperature","top_p"],"required_capabilities":["tools"],"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-4B-Instruct-2507::search","model_id":"Qwen/Qwen3-4B-Instruct-2507","canonical_id":"Qwen/Qwen3-4B-Instruct-2507","config_key":"search","variant_key":"search","interface":"huggingface","first_seen_at":"2026-04-13T19:43:59.804778+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-4B-Instruct-2507","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":951,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.306789+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-4B-Thinking-2507::base","model_id":"Qwen/Qwen3-4B-Thinking-2507","canonical_id":"Qwen/Qwen3-4B-Thinking-2507","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-4B-Thinking-2507","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":688,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.707705+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-4B-Thinking-2507::low_temp","model_id":"Qwen/Qwen3-4B-Thinking-2507","canonical_id":"Qwen/Qwen3-4B-Thinking-2507","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-4B-Thinking-2507","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":956,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.689041+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-4B-Instruct-2507::low_temp","model_id":"Qwen/Qwen3-4B-Instruct-2507","canonical_id":"Qwen/Qwen3-4B-Instruct-2507","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-4B-Instruct-2507","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":915,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.482104+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-7B-Instruct::structured","model_id":"Qwen/Qwen2.5-Coder-7B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-7B-Instruct","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-7B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1358,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.536675+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-4B-Instruct-2507::base","model_id":"Qwen/Qwen3-4B-Instruct-2507","canonical_id":"Qwen/Qwen3-4B-Instruct-2507","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-4B-Instruct-2507","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":4,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1006,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.502632+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-7B-Instruct::low_temp","model_id":"Qwen/Qwen2.5-Coder-7B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-7B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-7B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":950,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.501488+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-7B-Instruct::base","model_id":"Qwen/Qwen2.5-Coder-7B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-7B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-7B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1032,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.581749+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-3B-Instruct::structured","model_id":"Qwen/Qwen2.5-Coder-3B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-3B-Instruct","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":32768,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1410,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.794235+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-3B-Instruct::low_temp","model_id":"Qwen/Qwen2.5-Coder-3B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-3B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":32768,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:04:35.782319+00:00","last_latency_ms":1508,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:31.124078+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-3B-Instruct::base","model_id":"Qwen/Qwen2.5-Coder-3B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-3B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":32768,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.01,"pricing_output_per_1m":0.03,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:27:01.880407+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:04.969412+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::ibm-granite/granite-4.0-h-micro::base","model_id":"ibm-granite/granite-4.0-h-micro","canonical_id":"ibm-granite/granite-4.0-h-micro","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"IBM: Granite 4.0 Micro","org":"ibm-granite","org_name":"IBM Research","country":"USA","city":"Armonk, NY","org_type":"big-tech","open_weights":null,"context_length":131000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.017,"pricing_output_per_1m":0.11,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Granite series. Enterprise-focused. Strong code and business document understanding.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:14:50.016071+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::ibm-granite/granite-4.0-h-micro::low_temp","model_id":"ibm-granite/granite-4.0-h-micro","canonical_id":"ibm-granite/granite-4.0-h-micro","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"IBM: Granite 4.0 Micro","org":"ibm-granite","org_name":"IBM Research","country":"USA","city":"Armonk, NY","org_type":"big-tech","open_weights":null,"context_length":131000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.017,"pricing_output_per_1m":0.11,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Granite series. Enterprise-focused. Strong code and business document understanding.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:36.591563+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.1-8b-instruct::tools","model_id":"meta-llama/llama-3.1-8b-instruct","canonical_id":"meta-llama/llama-3.1-8b-instruct","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.1 8B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":16384,"max_completion_tokens":16384,"tokenizer":"Llama3","pricing_input_per_1m":0.02,"pricing_output_per_1m":0.049999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3","decoder_type":"Dense","attention":"GQA with RoPE","arch_highlight":"Pre-norm baseline; wider than OLMo 2 at a similar scale.","tech_report_url":"https://arxiv.org/pdf/2407.21783","hf_config_url":"https://huggingface.co/meta-llama/Meta-Llama-3-8B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:12.205724+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.1-8b-instruct::low_temp","model_id":"meta-llama/llama-3.1-8b-instruct","canonical_id":"meta-llama/llama-3.1-8b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.1 8B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":16384,"max_completion_tokens":16384,"tokenizer":"Llama3","pricing_input_per_1m":0.02,"pricing_output_per_1m":0.049999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3","decoder_type":"Dense","attention":"GQA with RoPE","arch_highlight":"Pre-norm baseline; wider than OLMo 2 at a similar scale.","tech_report_url":"https://arxiv.org/pdf/2407.21783","hf_config_url":"https://huggingface.co/meta-llama/Meta-Llama-3-8B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:39.307767+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.1-8b-instruct::structured","model_id":"meta-llama/llama-3.1-8b-instruct","canonical_id":"meta-llama/llama-3.1-8b-instruct","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.1 8B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":16384,"max_completion_tokens":16384,"tokenizer":"Llama3","pricing_input_per_1m":0.02,"pricing_output_per_1m":0.049999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":798,"first_unavailable_at":null,"arch_id":"llama-3","decoder_type":"Dense","attention":"GQA with RoPE","arch_highlight":"Pre-norm baseline; wider than OLMo 2 at a similar scale.","tech_report_url":"https://arxiv.org/pdf/2407.21783","hf_config_url":"https://huggingface.co/meta-llama/Meta-Llama-3-8B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.599827+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.1-8b-instruct::base","model_id":"meta-llama/llama-3.1-8b-instruct","canonical_id":"meta-llama/llama-3.1-8b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.1 8B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":16384,"max_completion_tokens":16384,"tokenizer":"Llama3","pricing_input_per_1m":0.02,"pricing_output_per_1m":0.049999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3","decoder_type":"Dense","attention":"GQA with RoPE","arch_highlight":"Pre-norm baseline; wider than OLMo 2 at a similar scale.","tech_report_url":"https://arxiv.org/pdf/2407.21783","hf_config_url":"https://huggingface.co/meta-llama/Meta-Llama-3-8B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T12:58:23.524464+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::meta-llama/Llama-3.1-8B-Instruct::structured","model_id":"meta-llama/Llama-3.1-8B-Instruct","canonical_id":"meta-llama/Llama-3.1-8B-Instruct","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Llama-3.1-8B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.02,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":374,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.521194+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":6},{"variant_id":"huggingface::meta-llama/Llama-3.1-8B-Instruct::search","model_id":"meta-llama/Llama-3.1-8B-Instruct","canonical_id":"meta-llama/Llama-3.1-8B-Instruct","config_key":"search","variant_key":"search","interface":"huggingface","first_seen_at":"2026-04-13T19:43:59.804778+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Llama-3.1-8B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.02,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":303,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.295591+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::meta-llama/Llama-3.1-8B-Instruct::low_temp","model_id":"meta-llama/Llama-3.1-8B-Instruct","canonical_id":"meta-llama/Llama-3.1-8B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Llama-3.1-8B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.02,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":312,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.471113+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":6},{"variant_id":"openrouter::mistralai/mistral-nemo::low_temp","model_id":"mistralai/mistral-nemo","canonical_id":"mistralai/mistral-nemo","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Nemo","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":16384,"tokenizer":"Mistral","pricing_input_per_1m":0.02,"pricing_output_per_1m":0.04,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:37.971265+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/mistral-nemo::base","model_id":"mistralai/mistral-nemo","canonical_id":"mistralai/mistral-nemo","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Nemo","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":16384,"tokenizer":"Mistral","pricing_input_per_1m":0.02,"pricing_output_per_1m":0.04,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:14:46.544629+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/mistral-nemo::structured","model_id":"mistralai/mistral-nemo","canonical_id":"mistralai/mistral-nemo","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Nemo","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":16384,"tokenizer":"Mistral","pricing_input_per_1m":0.02,"pricing_output_per_1m":0.04,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":378,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.601670+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::meta-llama/Llama-3.1-8B-Instruct::base","model_id":"meta-llama/Llama-3.1-8B-Instruct","canonical_id":"meta-llama/Llama-3.1-8B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Llama-3.1-8B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.02,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":662,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-16T01:13:43.316391+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":6},{"variant_id":"openrouter::mistralai/mistral-nemo::tools","model_id":"mistralai/mistral-nemo","canonical_id":"mistralai/mistral-nemo","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Nemo","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":16384,"tokenizer":"Mistral","pricing_input_per_1m":0.02,"pricing_output_per_1m":0.04,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:10.955995+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.2-1b-instruct::base","model_id":"meta-llama/llama-3.2-1b-instruct","canonical_id":"meta-llama/llama-3.2-1b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.2 1B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":60000,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.027,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":1,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3.2","decoder_type":"Dense","attention":"GQA","arch_highlight":"Reference small-model Llama architecture with tied embeddings.","tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:52:35.361049+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.2-1b-instruct::low_temp","model_id":"meta-llama/llama-3.2-1b-instruct","canonical_id":"meta-llama/llama-3.2-1b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.2 1B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":60000,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.027,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":1,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3.2","decoder_type":"Dense","attention":"GQA","arch_highlight":"Reference small-model Llama architecture with tied embeddings.","tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:41.543124+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3-8b-instruct::low_temp","model_id":"meta-llama/llama-3-8b-instruct","canonical_id":"meta-llama/llama-3-8b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3 8B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":8192,"max_completion_tokens":16384,"tokenizer":"Llama3","pricing_input_per_1m":0.03,"pricing_output_per_1m":0.04,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3","decoder_type":"Dense","attention":"GQA with RoPE","arch_highlight":"Pre-norm baseline; wider than OLMo 2 at a similar scale.","tech_report_url":"https://arxiv.org/pdf/2407.21783","hf_config_url":"https://huggingface.co/meta-llama/Meta-Llama-3-8B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:44.437283+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::liquid/lfm-2-24b-a2b-20260224::base","model_id":"liquid/lfm-2-24b-a2b","canonical_id":"liquid/lfm-2-24b-a2b-20260224","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"LiquidAI: LFM2-24B-A2B","org":"liquid","org_name":"Liquid AI","country":"USA","city":"Boston, MA","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.03,"pricing_output_per_1m":0.12,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":24,"active_param_count_b":2,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"LFM series. Non-transformer liquid foundation models (LFM). Novel architecture from MIT.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":213,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.990964+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::liquid/lfm-2-24b-a2b-20260224::search","model_id":"liquid/lfm-2-24b-a2b","canonical_id":"liquid/lfm-2-24b-a2b-20260224","config_key":"search","variant_key":"search","interface":"openrouter","first_seen_at":"2026-04-13T19:43:02.792642+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"LiquidAI: LFM2-24B-A2B","org":"liquid","org_name":"Liquid AI","country":"USA","city":"Boston, MA","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.03,"pricing_output_per_1m":0.12,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":24,"active_param_count_b":2,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"LFM series. Non-transformer liquid foundation models (LFM). Novel architecture from MIT.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":5229,"first_unavailable_at":"2026-04-13T20:35:56.784251+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.314297+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","stop","temperature","top_k","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::liquid/lfm-2-24b-a2b-20260224::low_temp","model_id":"liquid/lfm-2-24b-a2b","canonical_id":"liquid/lfm-2-24b-a2b-20260224","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"LiquidAI: LFM2-24B-A2B","org":"liquid","org_name":"Liquid AI","country":"USA","city":"Boston, MA","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.03,"pricing_output_per_1m":0.12,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":24,"active_param_count_b":2,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"LFM series. Non-transformer liquid foundation models (LFM). Novel architecture from MIT.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":3594,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.805267+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3-8b-instruct::tools","model_id":"meta-llama/llama-3-8b-instruct","canonical_id":"meta-llama/llama-3-8b-instruct","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3 8B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":8192,"max_completion_tokens":16384,"tokenizer":"Llama3","pricing_input_per_1m":0.03,"pricing_output_per_1m":0.04,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-03-15T19:50:01.835000+00:00","arch_id":"llama-3","decoder_type":"Dense","attention":"GQA with RoPE","arch_highlight":"Pre-norm baseline; wider than OLMo 2 at a similar scale.","tech_report_url":"https://arxiv.org/pdf/2407.21783","hf_config_url":"https://huggingface.co/meta-llama/Meta-Llama-3-8B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:13.417908+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3-8b-instruct::base","model_id":"meta-llama/llama-3-8b-instruct","canonical_id":"meta-llama/llama-3-8b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3 8B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":8192,"max_completion_tokens":16384,"tokenizer":"Llama3","pricing_input_per_1m":0.03,"pricing_output_per_1m":0.04,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"llama-3","decoder_type":"Dense","attention":"GQA with RoPE","arch_highlight":"Pre-norm baseline; wider than OLMo 2 at a similar scale.","tech_report_url":"https://arxiv.org/pdf/2407.21783","hf_config_url":"https://huggingface.co/meta-llama/Meta-Llama-3-8B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:12:46.987455+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::amazon/nova-micro-v1::tools","model_id":"amazon/nova-micro-v1","canonical_id":"amazon/nova-micro-v1","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Amazon: Nova Micro 1.0","org":"amazon","org_name":"Amazon / AWS","country":"USA","city":"Seattle, WA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":5120,"tokenizer":"Nova","pricing_input_per_1m":0.035,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","stop","temperature","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nova series. Amazon's proprietary LLMs. Heavy e-commerce and AWS domain data.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:14.976274+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","stop","temperature","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::amazon/nova-micro-v1::low_temp","model_id":"amazon/nova-micro-v1","canonical_id":"amazon/nova-micro-v1","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Amazon: Nova Micro 1.0","org":"amazon","org_name":"Amazon / AWS","country":"USA","city":"Seattle, WA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":5120,"tokenizer":"Nova","pricing_input_per_1m":0.035,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","stop","temperature","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nova series. Amazon's proprietary LLMs. Heavy e-commerce and AWS domain data.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:45.901222+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","stop","temperature","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::amazon/nova-micro-v1::base","model_id":"amazon/nova-micro-v1","canonical_id":"amazon/nova-micro-v1","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Amazon: Nova Micro 1.0","org":"amazon","org_name":"Amazon / AWS","country":"USA","city":"Seattle, WA","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":5120,"tokenizer":"Nova","pricing_input_per_1m":0.035,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","stop","temperature","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nova series. Amazon's proprietary LLMs. Heavy e-commerce and AWS domain data.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:52:41.682152+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","stop","temperature","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::cohere/command-r7b-12-2024::low_temp","model_id":"cohere/command-r7b-12-2024","canonical_id":"cohere/command-r7b-12-2024","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Cohere: Command R7B (12-2024)","org":"cohere","org_name":"Cohere","country":"Canada","city":"Toronto, ON","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":4000,"tokenizer":"Cohere","pricing_input_per_1m":0.0375,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Command series. Enterprise-focused. Strong RAG and retrieval performance. Canadian AI lab.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:47.724051+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::cohere/command-r7b-12-2024::structured","model_id":"cohere/command-r7b-12-2024","canonical_id":"cohere/command-r7b-12-2024","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Cohere: Command R7B (12-2024)","org":"cohere","org_name":"Cohere","country":"Canada","city":"Toronto, ON","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":4000,"tokenizer":"Cohere","pricing_input_per_1m":0.0375,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Command series. Enterprise-focused. Strong RAG and retrieval performance. Canadian AI lab.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":449,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.511255+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::cohere/command-r7b-12-2024::base","model_id":"cohere/command-r7b-12-2024","canonical_id":"cohere/command-r7b-12-2024","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Cohere: Command R7B (12-2024)","org":"cohere","org_name":"Cohere","country":"Canada","city":"Toronto, ON","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":4000,"tokenizer":"Cohere","pricing_input_per_1m":0.0375,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Command series. Enterprise-focused. Strong RAG and retrieval performance. Canadian AI lab.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:14.133340+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-oss-20b::structured","model_id":"openai/gpt-oss-20b","canonical_id":"openai/gpt-oss-20b","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-20b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":348,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.215094+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::sao10k/l3-lunaris-8b::low_temp","model_id":"sao10k/l3-lunaris-8b","canonical_id":"sao10k/l3-lunaris-8b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Sao10K: Llama 3 8B Lunaris","org":"sao10k","org_name":"Sao10k (community)","country":"Unknown","city":"Unknown","org_type":"community","open_weights":null,"context_length":8192,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.049999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Community fine-tunes of Llama for creative/roleplay use cases.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:50.563343+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::sao10k/l3-lunaris-8b::base","model_id":"sao10k/l3-lunaris-8b","canonical_id":"sao10k/l3-lunaris-8b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Sao10K: Llama 3 8B Lunaris","org":"sao10k","org_name":"Sao10k (community)","country":"Unknown","city":"Unknown","org_type":"community","open_weights":null,"context_length":8192,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.049999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Community fine-tunes of Llama for creative/roleplay use cases.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:15.514040+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::meta-llama/Meta-Llama-3-8B-Instruct::search","model_id":"meta-llama/Meta-Llama-3-8B-Instruct","canonical_id":"meta-llama/Meta-Llama-3-8B-Instruct","config_key":"search","variant_key":"search","interface":"huggingface","first_seen_at":"2026-04-13T19:43:59.804778+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta-Llama-3-8B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.04,"pricing_output_per_1m":0.04,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":690,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.300863+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::meta-llama/Meta-Llama-3-8B-Instruct::low_temp","model_id":"meta-llama/Meta-Llama-3-8B-Instruct","canonical_id":"meta-llama/Meta-Llama-3-8B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta-Llama-3-8B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.04,"pricing_output_per_1m":0.04,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":597,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.476950+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"openrouter::qwen/qwen-2.5-7b-instruct::tools","model_id":"qwen/qwen-2.5-7b-instruct","canonical_id":"qwen/qwen-2.5-7b-instruct","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen2.5 7B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":32768,"tokenizer":"Qwen","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-03-15T19:50:01.835000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:16.580411+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::sao10k/l3-lunaris-8b::structured","model_id":"sao10k/l3-lunaris-8b","canonical_id":"sao10k/l3-lunaris-8b","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Sao10K: Llama 3 8B Lunaris","org":"sao10k","org_name":"Sao10k (community)","country":"Unknown","city":"Unknown","org_type":"community","open_weights":null,"context_length":8192,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.049999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Community fine-tunes of Llama for creative/roleplay use cases.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":4029,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.596021+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::meta-llama/Meta-Llama-3-8B-Instruct::base","model_id":"meta-llama/Meta-Llama-3-8B-Instruct","canonical_id":"meta-llama/Meta-Llama-3-8B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta-Llama-3-8B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.04,"pricing_output_per_1m":0.04,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":720,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.498567+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"openrouter::qwen/qwen-2.5-7b-instruct::low_temp","model_id":"qwen/qwen-2.5-7b-instruct","canonical_id":"qwen/qwen-2.5-7b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen2.5 7B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":32768,"tokenizer":"Qwen","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:51.965339+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::meta-llama/Meta-Llama-3-8B-Instruct::structured","model_id":"meta-llama/Meta-Llama-3-8B-Instruct","canonical_id":"meta-llama/Meta-Llama-3-8B-Instruct","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-06T06:45:19.840772+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta-Llama-3-8B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.04,"pricing_output_per_1m":0.04,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":752,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.786889+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":2},{"variant_id":"huggingface::openai/gpt-oss-20b::search","model_id":"openai/gpt-oss-20b","canonical_id":"openai/gpt-oss-20b","config_key":"search","variant_key":"search","interface":"huggingface","first_seen_at":"2026-04-13T19:43:59.804778+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-20b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":216,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.296655+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::openai/gpt-oss-20b::low_temp","model_id":"openai/gpt-oss-20b","canonical_id":"openai/gpt-oss-20b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-20b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":241,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.469420+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":3},{"variant_id":"huggingface::openai/gpt-oss-20b::base","model_id":"openai/gpt-oss-20b","canonical_id":"openai/gpt-oss-20b","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-20b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":20,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":233,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.488523+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":3},{"variant_id":"openrouter::qwen/qwen-2.5-7b-instruct::base","model_id":"qwen/qwen-2.5-7b-instruct","canonical_id":"qwen/qwen-2.5-7b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen2.5 7B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":32768,"tokenizer":"Qwen","pricing_input_per_1m":0.04,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:14:39.256355+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-24b-instruct-2501::low_temp","model_id":"mistralai/mistral-small-24b-instruct-2501","canonical_id":"mistralai/mistral-small-24b-instruct-2501","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 3","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":16384,"tokenizer":"Mistral","pricing_input_per_1m":0.049999999999999996,"pricing_output_per_1m":0.08,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:54.053758+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-5-nano-2025-08-07::tools+reasoning","model_id":"openai/gpt-5-nano","canonical_id":"openai/gpt-5-nano-2025-08-07","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-5 Nano","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":400000,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.049999999999999996,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_completion_tokens","max_tokens","reasoning","response_format","seed","structured_outputs","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":3390,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.390060+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_completion_tokens","max_tokens","reasoning","response_format","seed","structured_outputs","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::openai/gpt-5-nano-2025-08-07::reasoning","model_id":"openai/gpt-5-nano","canonical_id":"openai/gpt-5-nano-2025-08-07","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-5 Nano","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":400000,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.049999999999999996,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_completion_tokens","max_tokens","reasoning","response_format","seed","structured_outputs","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2818,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.712757+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_completion_tokens","max_tokens","reasoning","response_format","seed","structured_outputs","tool_choice","tools"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-24b-instruct-2501::base","model_id":"mistralai/mistral-small-24b-instruct-2501","canonical_id":"mistralai/mistral-small-24b-instruct-2501","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 3","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":16384,"tokenizer":"Mistral","pricing_input_per_1m":0.049999999999999996,"pricing_output_per_1m":0.08,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:10:49.154533+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-8b-04-28::reasoning","model_id":"qwen/qwen3-8b","canonical_id":"qwen/qwen3-8b-04-28","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 8B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":8192,"tokenizer":"Qwen3","pricing_input_per_1m":0.049999999999999996,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2497,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":"qwen3-dense","decoder_type":"Dense","attention":"GQA with QK-Norm","arch_highlight":"Reference dense Qwen stack with QK-Norm and 8 KV heads.","tech_report_url":"https://arxiv.org/pdf/2505.09388","hf_config_url":"https://huggingface.co/Qwen/Qwen3-32B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.821290+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-8b-04-28::structured","model_id":"qwen/qwen3-8b","canonical_id":"qwen/qwen3-8b-04-28","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 8B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":8192,"tokenizer":"Qwen3","pricing_input_per_1m":0.049999999999999996,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":10613,"first_unavailable_at":"2026-03-15T19:48:48.939000+00:00","arch_id":"qwen3-dense","decoder_type":"Dense","attention":"GQA with QK-Norm","arch_highlight":"Reference dense Qwen stack with QK-Norm and 8 KV heads.","tech_report_url":"https://arxiv.org/pdf/2505.09388","hf_config_url":"https://huggingface.co/Qwen/Qwen3-32B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.387432+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::openai/gpt-oss-120b::search","model_id":"openai/gpt-oss-120b","canonical_id":"openai/gpt-oss-120b","config_key":"search","variant_key":"search","interface":"huggingface","first_seen_at":"2026-04-13T19:43:59.804778+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-120b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.05,"pricing_output_per_1m":0.25,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":193,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.292610+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Sao10K/L3-8B-Stheno-v3.2::base","model_id":"Sao10K/L3-8B-Stheno-v3.2","canonical_id":"Sao10K/L3-8B-Stheno-v3.2","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"L3-8B-Stheno-v3.2","org":"Sao10K","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.05,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":692,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.680635+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Sao10K/L3-8B-Lunaris-v1::low_temp","model_id":"Sao10K/L3-8B-Lunaris-v1","canonical_id":"Sao10K/L3-8B-Lunaris-v1","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"L3-8B-Lunaris-v1","org":"Sao10K","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.05,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:03:43.384368+00:00","last_latency_ms":884,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:57.322455+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"huggingface::deepseek-ai/DeepSeek-R1-Distill-Llama-8B::low_temp","model_id":"deepseek-ai/DeepSeek-R1-Distill-Llama-8B","canonical_id":"deepseek-ai/DeepSeek-R1-Distill-Llama-8B","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek-R1-Distill-Llama-8B","org":"deepseek-ai","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.05,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":667,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.606191+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::deepseek-ai/DeepSeek-R1-Distill-Llama-8B::base","model_id":"deepseek-ai/DeepSeek-R1-Distill-Llama-8B","canonical_id":"deepseek-ai/DeepSeek-R1-Distill-Llama-8B","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek-R1-Distill-Llama-8B","org":"deepseek-ai","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.05,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1085,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.694789+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Sao10K/L3-8B-Lunaris-v1::base","model_id":"Sao10K/L3-8B-Lunaris-v1","canonical_id":"Sao10K/L3-8B-Lunaris-v1","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"L3-8B-Lunaris-v1","org":"Sao10K","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.05,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:27:01.880407+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:11:06.107460+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"huggingface::Sao10K/L3-8B-Stheno-v3.2::low_temp","model_id":"Sao10K/L3-8B-Stheno-v3.2","canonical_id":"Sao10K/L3-8B-Stheno-v3.2","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"L3-8B-Stheno-v3.2","org":"Sao10K","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.05,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":687,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.596459+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"openrouter::openai/gpt-oss-120b::structured","model_id":"openai/gpt-oss-120b","canonical_id":"openai/gpt-oss-120b","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-120b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.05,"pricing_output_per_1m":0.25,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":257,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.213227+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::openai/gpt-oss-120b::tools","model_id":"openai/gpt-oss-120b","canonical_id":"openai/gpt-oss-120b","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-120b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.05,"pricing_output_per_1m":0.25,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:17.741288+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":3},{"variant_id":"huggingface::openai/gpt-oss-120b::structured","model_id":"openai/gpt-oss-120b","canonical_id":"openai/gpt-oss-120b","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-120b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.05,"pricing_output_per_1m":0.25,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":342,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.785215+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":3},{"variant_id":"huggingface::openai/gpt-oss-120b::low_temp","model_id":"openai/gpt-oss-120b","canonical_id":"openai/gpt-oss-120b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-120b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.05,"pricing_output_per_1m":0.25,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:42:58.482734+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":3},{"variant_id":"huggingface::openai/gpt-oss-120b::base","model_id":"openai/gpt-oss-120b","canonical_id":"openai/gpt-oss-120b","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"gpt-oss-120b","org":"openai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":"GPT","pricing_input_per_1m":0.05,"pricing_output_per_1m":0.25,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":120,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:21.775509+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":3},{"variant_id":"huggingface::deepseek-ai/DeepSeek-R1-Distill-Llama-8B::structured","model_id":"deepseek-ai/DeepSeek-R1-Distill-Llama-8B","canonical_id":"deepseek-ai/DeepSeek-R1-Distill-Llama-8B","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-06T06:45:19.840772+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek-R1-Distill-Llama-8B","org":"deepseek-ai","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.05,"pricing_output_per_1m":0.05,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":2183,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.788802+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":1},{"variant_id":"openrouter::qwen/qwen3-14b-04-28::reasoning","model_id":"qwen/qwen3-14b","canonical_id":"qwen/qwen3-14b-04-28","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 14B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":40960,"tokenizer":"Qwen3","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.24,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":14,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1084,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.882033+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.7-flash-20260119::reasoning","model_id":"z-ai/glm-4.7-flash","canonical_id":"z-ai/glm-4.7-flash-20260119","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.7 Flash","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":202752,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1005,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.486369+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.7-flash-20260119::low_temp","model_id":"z-ai/glm-4.7-flash","canonical_id":"z-ai/glm-4.7-flash-20260119","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.7 Flash","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":202752,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":210,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.897737+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.7-flash-20260119::structured","model_id":"z-ai/glm-4.7-flash","canonical_id":"z-ai/glm-4.7-flash-20260119","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.7 Flash","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":202752,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":756,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.888734+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::amazon/nova-lite-v1::tools","model_id":"amazon/nova-lite-v1","canonical_id":"amazon/nova-lite-v1","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Amazon: Nova Lite 1.0","org":"amazon","org_name":"Amazon / AWS","country":"USA","city":"Seattle, WA","org_type":"big-tech","open_weights":null,"context_length":300000,"max_completion_tokens":5120,"tokenizer":"Nova","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.24,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","stop","temperature","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nova series. Amazon's proprietary LLMs. Heavy e-commerce and AWS domain data.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:19.650950+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","stop","temperature","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.7-flash-20260119::base","model_id":"z-ai/glm-4.7-flash","canonical_id":"z-ai/glm-4.7-flash-20260119","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.7 Flash","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":202752,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":283,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.107450+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-32B-Instruct::structured","model_id":"Qwen/Qwen2.5-Coder-32B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-32B-Instruct","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-32B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.06,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":500,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.608681+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":2},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-32B-Instruct::low_temp","model_id":"Qwen/Qwen2.5-Coder-32B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-32B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-32B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.06,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":862,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.794711+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"huggingface::Qwen/Qwen2.5-Coder-32B-Instruct::base","model_id":"Qwen/Qwen2.5-Coder-32B-Instruct","canonical_id":"Qwen/Qwen2.5-Coder-32B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5-Coder-32B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.06,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":285,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.807245+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"openrouter::gryphe/mythomax-l2-13b::low_temp","model_id":"gryphe/mythomax-l2-13b","canonical_id":"gryphe/mythomax-l2-13b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MythoMax 13B","org":"gryphe","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":4096,"max_completion_tokens":4096,"tokenizer":"Llama2","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.06,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":13,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_a","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:43:05.982643+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_a","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::gryphe/mythomax-l2-13b::base","model_id":"gryphe/mythomax-l2-13b","canonical_id":"gryphe/mythomax-l2-13b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MythoMax 13B","org":"gryphe","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":4096,"max_completion_tokens":4096,"tokenizer":"Llama2","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.06,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":13,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_a","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:11:21.338903+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_a","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::amazon/nova-lite-v1::low_temp","model_id":"amazon/nova-lite-v1","canonical_id":"amazon/nova-lite-v1","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Amazon: Nova Lite 1.0","org":"amazon","org_name":"Amazon / AWS","country":"USA","city":"Seattle, WA","org_type":"big-tech","open_weights":null,"context_length":300000,"max_completion_tokens":5120,"tokenizer":"Nova","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.24,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","stop","temperature","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nova series. Amazon's proprietary LLMs. Heavy e-commerce and AWS domain data.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:43:07.309493+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","stop","temperature","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::amazon/nova-lite-v1::base","model_id":"amazon/nova-lite-v1","canonical_id":"amazon/nova-lite-v1","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Amazon: Nova Lite 1.0","org":"amazon","org_name":"Amazon / AWS","country":"USA","city":"Seattle, WA","org_type":"big-tech","open_weights":null,"context_length":300000,"max_completion_tokens":5120,"tokenizer":"Nova","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.24,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","stop","temperature","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nova series. Amazon's proprietary LLMs. Heavy e-commerce and AWS domain data.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:15:34.184262+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","stop","temperature","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.7-flash-20260119::tools+reasoning","model_id":"z-ai/glm-4.7-flash","canonical_id":"z-ai/glm-4.7-flash-20260119","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.7 Flash","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":202752,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.06,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":2352,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.089791+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-flash-20260224::base","model_id":"qwen/qwen3.5-flash-02-23","canonical_id":"qwen/qwen3.5-flash-20260224","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-Flash","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.065,"pricing_output_per_1m":0.26,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":7367,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.989617+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-flash-20260224::low_temp","model_id":"qwen/qwen3.5-flash-02-23","canonical_id":"qwen/qwen3.5-flash-20260224","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-Flash","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.065,"pricing_output_per_1m":0.26,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":5102,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.803265+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-flash-20260224::reasoning","model_id":"qwen/qwen3.5-flash-02-23","canonical_id":"qwen/qwen3.5-flash-20260224","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-Flash","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.065,"pricing_output_per_1m":0.26,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2523,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.403879+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::microsoft/phi-4::structured","model_id":"microsoft/phi-4","canonical_id":"microsoft/phi-4","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Microsoft: Phi 4","org":"microsoft","org_name":"Microsoft Research","country":"USA","city":"Redmond, WA","org_type":"big-tech","open_weights":null,"context_length":16384,"max_completion_tokens":16384,"tokenizer":"Other","pricing_input_per_1m":0.065,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Phi series. Textbook-quality synthetic training data. Small but capable. WizardLM instruction fine-tunes.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":521,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.505749+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::microsoft/phi-4::base","model_id":"microsoft/phi-4","canonical_id":"microsoft/phi-4","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Microsoft: Phi 4","org":"microsoft","org_name":"Microsoft Research","country":"USA","city":"Redmond, WA","org_type":"big-tech","open_weights":null,"context_length":16384,"max_completion_tokens":16384,"tokenizer":"Other","pricing_input_per_1m":0.065,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Phi series. Textbook-quality synthetic training data. Small but capable. WizardLM instruction fine-tunes.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:33.265728+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::microsoft/phi-4::low_temp","model_id":"microsoft/phi-4","canonical_id":"microsoft/phi-4","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Microsoft: Phi 4","org":"microsoft","org_name":"Microsoft Research","country":"USA","city":"Redmond, WA","org_type":"big-tech","open_weights":null,"context_length":16384,"max_completion_tokens":16384,"tokenizer":"Other","pricing_input_per_1m":0.065,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Phi series. Textbook-quality synthetic training data. Small but capable. WizardLM instruction fine-tunes.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:43:08.848712+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-Coder-30B-A3B-Instruct::structured","model_id":"Qwen/Qwen3-Coder-30B-A3B-Instruct","canonical_id":"Qwen/Qwen3-Coder-30B-A3B-Instruct","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Coder-30B-A3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.26,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":510,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.532670+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":3},{"variant_id":"huggingface::Qwen/Qwen3-Coder-30B-A3B-Instruct::low_temp","model_id":"Qwen/Qwen3-Coder-30B-A3B-Instruct","canonical_id":"Qwen/Qwen3-Coder-30B-A3B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Coder-30B-A3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.26,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":468,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.494038+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":3},{"variant_id":"huggingface::zai-org/GLM-4.7-Flash::tools","model_id":"zai-org/GLM-4.7-Flash","canonical_id":"zai-org/GLM-4.7-Flash","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-08T01:41:35.453255+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"GLM-4.7-Flash","org":"zai-org","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":200000,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.4,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-08T01:41:35.453255+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:39.096995+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":2},{"variant_id":"huggingface::Qwen/Qwen3-Coder-30B-A3B-Instruct::base","model_id":"Qwen/Qwen3-Coder-30B-A3B-Instruct","canonical_id":"Qwen/Qwen3-Coder-30B-A3B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Coder-30B-A3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.26,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":494,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.513677+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":3},{"variant_id":"huggingface::Qwen/Qwen3-8B::tools","model_id":"Qwen/Qwen3-8B","canonical_id":"Qwen/Qwen3-8B","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-8B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.18,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:10:18.669859+00:00","last_latency_ms":470,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:21.174839+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":2},{"variant_id":"huggingface::Qwen/Qwen3-8B::search","model_id":"Qwen/Qwen3-8B","canonical_id":"Qwen/Qwen3-8B","config_key":"search","variant_key":"search","interface":"huggingface","first_seen_at":"2026-04-13T19:43:59.804778+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-8B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.18,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":411,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.305516+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-8B::low_temp","model_id":"Qwen/Qwen3-8B","canonical_id":"Qwen/Qwen3-8B","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-8B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.18,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":366,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.478085+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"huggingface::Qwen/Qwen3-8B::base","model_id":"Qwen/Qwen3-8B","canonical_id":"Qwen/Qwen3-8B","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-8B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.18,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1652,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.889468+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"openrouter::baidu/ernie-4.5-21b-a3b::base","model_id":"baidu/ernie-4.5-21b-a3b","canonical_id":"baidu/ernie-4.5-21b-a3b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Baidu: ERNIE 4.5 21B A3B","org":"baidu","org_name":"Baidu","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":120000,"max_completion_tokens":8000,"tokenizer":"Other","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":21,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"ERNIE series. Dominant Chinese search/web corpus. Strong Chinese language and cultural knowledge.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:55:24.280224+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-14B::structured","model_id":"Qwen/Qwen3-14B","canonical_id":"Qwen/Qwen3-14B","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-06T06:45:19.840772+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-14B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":14,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":3185,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.713033+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":1},{"variant_id":"openrouter::baidu/ernie-4.5-21b-a3b::tools","model_id":"baidu/ernie-4.5-21b-a3b","canonical_id":"baidu/ernie-4.5-21b-a3b","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Baidu: ERNIE 4.5 21B A3B","org":"baidu","org_name":"Baidu","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":120000,"max_completion_tokens":8000,"tokenizer":"Other","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":21,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"ERNIE series. Dominant Chinese search/web corpus. Strong Chinese language and cultural knowledge.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:43.698323+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::baidu/ernie-4.5-21b-a3b-thinking::reasoning","model_id":"baidu/ernie-4.5-21b-a3b-thinking","canonical_id":"baidu/ernie-4.5-21b-a3b-thinking","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Baidu: ERNIE 4.5 21B A3B Thinking","org":"baidu","org_name":"Baidu","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":21,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"ERNIE series. Dominant Chinese search/web corpus. Strong Chinese language and cultural knowledge.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2092,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.611054+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","seed","stop","temperature","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::baidu/ernie-4.5-21b-a3b-thinking::low_temp","model_id":"baidu/ernie-4.5-21b-a3b-thinking","canonical_id":"baidu/ernie-4.5-21b-a3b-thinking","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Baidu: ERNIE 4.5 21B A3B Thinking","org":"baidu","org_name":"Baidu","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":21,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"ERNIE series. Dominant Chinese search/web corpus. Strong Chinese language and cultural knowledge.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:43:49.579993+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","seed","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::baidu/ernie-4.5-21b-a3b-thinking::base","model_id":"baidu/ernie-4.5-21b-a3b-thinking","canonical_id":"baidu/ernie-4.5-21b-a3b-thinking","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Baidu: ERNIE 4.5 21B A3B Thinking","org":"baidu","org_name":"Baidu","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":21,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"ERNIE series. Dominant Chinese search/web corpus. Strong Chinese language and cultural knowledge.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:15:00.075894+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","seed","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::baidu/ernie-4.5-21b-a3b::low_temp","model_id":"baidu/ernie-4.5-21b-a3b","canonical_id":"baidu/ernie-4.5-21b-a3b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Baidu: ERNIE 4.5 21B A3B","org":"baidu","org_name":"Baidu","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":120000,"max_completion_tokens":8000,"tokenizer":"Other","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":21,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"ERNIE series. Dominant Chinese search/web corpus. Strong Chinese language and cultural knowledge.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:18.711446+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-14B::low_temp","model_id":"Qwen/Qwen3-14B","canonical_id":"Qwen/Qwen3-14B","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-14B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":14,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":995,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.680980+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-14B::base","model_id":"Qwen/Qwen3-14B","canonical_id":"Qwen/Qwen3-14B","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-14B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":14,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:41.291454+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"openrouter::qwen/qwen3-coder-30b-a3b-instruct::low_temp","model_id":"qwen/qwen3-coder-30b-a3b-instruct","canonical_id":"qwen/qwen3-coder-30b-a3b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder 30B A3B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":160000,"max_completion_tokens":32768,"tokenizer":"Qwen3","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.27,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:43:15.132971+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-30b-a3b-instruct::structured","model_id":"qwen/qwen3-coder-30b-a3b-instruct","canonical_id":"qwen/qwen3-coder-30b-a3b-instruct","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder 30B A3B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":160000,"max_completion_tokens":32768,"tokenizer":"Qwen3","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.27,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":579,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.220585+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-30b-a3b-instruct::base","model_id":"qwen/qwen3-coder-30b-a3b-instruct","canonical_id":"qwen/qwen3-coder-30b-a3b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder 30B A3B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":160000,"max_completion_tokens":32768,"tokenizer":"Qwen3","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.27,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:34.656965+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-30b-a3b-instruct::tools","model_id":"qwen/qwen3-coder-30b-a3b-instruct","canonical_id":"qwen/qwen3-coder-30b-a3b-instruct","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder 30B A3B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":160000,"max_completion_tokens":32768,"tokenizer":"Qwen3","pricing_input_per_1m":0.07,"pricing_output_per_1m":0.27,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:40.989363+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"huggingface::baidu/ERNIE-4.5-21B-A3B-PT::base","model_id":"baidu/ERNIE-4.5-21B-A3B-PT","canonical_id":"baidu/ERNIE-4.5-21B-A3B-PT","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"ERNIE-4.5-21B-A3B-PT","org":"baidu","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":120000,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.07,"pricing_output_per_1m":0.28,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":21,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1328,"first_unavailable_at":"2026-04-12T07:44:48.885180+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.703646+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"openrouter::qwen/qwen3-235b-a22b-07-25::structured","model_id":"qwen/qwen3-235b-a22b-2507","canonical_id":"qwen/qwen3-235b-a22b-07-25","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 235B A22B Instruct 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0.071,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":3807,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.281621+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-235b-a22b-07-25::reasoning","model_id":"qwen/qwen3-235b-a22b-2507","canonical_id":"qwen/qwen3-235b-a22b-07-25","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 235B A22B Instruct 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0.071,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":266,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.791645+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-235b-a22b-07-25::tools+reasoning","model_id":"qwen/qwen3-235b-a22b-2507","canonical_id":"qwen/qwen3-235b-a22b-07-25","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 235B A22B Instruct 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0.071,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1118,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.480786+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.0-flash-lite-001::structured","model_id":"google/gemini-2.0-flash-lite-001","canonical_id":"google/gemini-2.0-flash-lite-001","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.0 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":"2026-06-01","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":967,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.808268+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.0-flash-lite-001::tools","model_id":"google/gemini-2.0-flash-lite-001","canonical_id":"google/gemini-2.0-flash-lite-001","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.0 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":"2026-06-01","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:45.082288+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.0-flash-lite-001::base","model_id":"google/gemini-2.0-flash-lite-001","canonical_id":"google/gemini-2.0-flash-lite-001","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.0 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":"2026-06-01","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:22:53.086255+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:43.089476+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-3.2-24b-instruct-2506::structured","model_id":"mistralai/mistral-small-3.2-24b-instruct","canonical_id":"mistralai/mistral-small-3.2-24b-instruct-2506","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 3.2 24B","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":24,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":413,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.299825+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-1.6-flash-20250625::tools+reasoning","model_id":"bytedance-seed/seed-1.6-flash","canonical_id":"bytedance-seed/seed-1.6-flash-20250625","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed 1.6 Flash","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":642,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.094489+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-1.6-flash-20250625::structured","model_id":"bytedance-seed/seed-1.6-flash","canonical_id":"bytedance-seed/seed-1.6-flash-20250625","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed 1.6 Flash","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1553,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.893875+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-1.6-flash-20250625::reasoning","model_id":"bytedance-seed/seed-1.6-flash","canonical_id":"bytedance-seed/seed-1.6-flash-20250625","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed 1.6 Flash","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1472,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.490035+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-1.6-flash-20250625::low_temp","model_id":"bytedance-seed/seed-1.6-flash","canonical_id":"bytedance-seed/seed-1.6-flash-20250625","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed 1.6 Flash","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":906,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.901322+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-1.6-flash-20250625::base","model_id":"bytedance-seed/seed-1.6-flash","canonical_id":"bytedance-seed/seed-1.6-flash-20250625","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed 1.6 Flash","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":32768,"tokenizer":"Other","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1090,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.113045+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.0-flash-lite-001::low_temp","model_id":"google/gemini-2.0-flash-lite-001","canonical_id":"google/gemini-2.0-flash-lite-001","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.0 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0.075,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":"2026-06-01","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:06.384978+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:20.981987+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemma-3-27b-it::structured","model_id":"google/gemma-3-27b-it","canonical_id":"google/gemma-3-27b-it","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 3 27B","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":16384,"tokenizer":"Gemini","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.16,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":417,"first_unavailable_at":null,"arch_id":"gemma-3","decoder_type":"Dense","attention":"GQA with QK-Norm and 5:1 sliding-window/global attention","arch_highlight":"Built around a 27B sweet spot with heavier local attention and a large 262k multilingual vocabulary.","tech_report_url":"https://arxiv.org/pdf/2503.19786","hf_config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.494534+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-32b-04-28::tools+reasoning","model_id":"qwen/qwen3-32b","canonical_id":"qwen/qwen3-32b-04-28","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 32B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":40960,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.24,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":2028,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":"qwen3-dense","decoder_type":"Dense","attention":"GQA with QK-Norm","arch_highlight":"Reference dense Qwen stack with QK-Norm and 8 KV heads.","tech_report_url":"https://arxiv.org/pdf/2505.09388","hf_config_url":"https://huggingface.co/Qwen/Qwen3-32B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.596882+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-04-28::tools+reasoning","model_id":"qwen/qwen3-30b-a3b","canonical_id":"qwen/qwen3-30b-a3b-04-28","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":40960,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":3478,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":"qwen3-moe","decoder_type":"Sparse MoE","attention":"GQA with QK-Norm","arch_highlight":"High-capacity MoE design optimized for serving efficiency without a shared expert.","tech_report_url":"https://arxiv.org/pdf/2505.09388","hf_config_url":"https://huggingface.co/Qwen/Qwen3-235B-A22B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.589804+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-thinking-2507::tools+reasoning","model_id":"qwen/qwen3-30b-a3b-thinking-2507","canonical_id":"qwen/qwen3-30b-a3b-thinking-2507","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":3409,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.310368+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-thinking-2507::structured","model_id":"qwen/qwen3-30b-a3b-thinking-2507","canonical_id":"qwen/qwen3-30b-a3b-thinking-2507","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":3609,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.716876+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-thinking-2507::low_temp","model_id":"qwen/qwen3-30b-a3b-thinking-2507","canonical_id":"qwen/qwen3-30b-a3b-thinking-2507","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:30.163624+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-thinking-2507::base","model_id":"qwen/qwen3-30b-a3b-thinking-2507","canonical_id":"qwen/qwen3-30b-a3b-thinking-2507","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:46.416956+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-32b-04-28::reasoning","model_id":"qwen/qwen3-32b","canonical_id":"qwen/qwen3-32b-04-28","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 32B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":40960,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.24,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":6203,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":"qwen3-dense","decoder_type":"Dense","attention":"GQA with QK-Norm","arch_highlight":"Reference dense Qwen stack with QK-Norm and 8 KV heads.","tech_report_url":"https://arxiv.org/pdf/2505.09388","hf_config_url":"https://huggingface.co/Qwen/Qwen3-32B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.883695+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-32b-04-28::structured","model_id":"qwen/qwen3-32b","canonical_id":"qwen/qwen3-32b-04-28","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 32B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":40960,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.24,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":5028,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":"qwen3-dense","decoder_type":"Dense","attention":"GQA with QK-Norm","arch_highlight":"Reference dense Qwen stack with QK-Norm and 8 KV heads.","tech_report_url":"https://arxiv.org/pdf/2505.09388","hf_config_url":"https://huggingface.co/Qwen/Qwen3-32B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.391930+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-04-28::reasoning","model_id":"qwen/qwen3-30b-a3b","canonical_id":"qwen/qwen3-30b-a3b-04-28","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":40960,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1483,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":"qwen3-moe","decoder_type":"Sparse MoE","attention":"GQA with QK-Norm","arch_highlight":"High-capacity MoE design optimized for serving efficiency without a shared expert.","tech_report_url":"https://arxiv.org/pdf/2505.09388","hf_config_url":"https://huggingface.co/Qwen/Qwen3-235B-A22B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.819584+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-thinking-2507::tools","model_id":"qwen/qwen3-30b-a3b-thinking-2507","canonical_id":"qwen/qwen3-30b-a3b-thinking-2507","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:47.802238+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-32B::low_temp","model_id":"Qwen/Qwen3-32B","canonical_id":"Qwen/Qwen3-32B","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-32B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.08,"pricing_output_per_1m":0.25,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":591,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.590221+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":4},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-thinking-2507::reasoning","model_id":"qwen/qwen3-30b-a3b-thinking-2507","canonical_id":"qwen/qwen3-30b-a3b-thinking-2507","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2923,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.696699+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::google/gemma-4-26b-a4b-it-20260403::structured","model_id":"google/gemma-4-26b-a4b-it","canonical_id":"google/gemma-4-26b-a4b-it-20260403","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-04-06T06:45:01.382737+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 26B A4B ","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Gemma","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.35,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":26,"active_param_count_b":4,"is_moe":true,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-06T06:45:01.382737+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":790,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.695618+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemma-4-26b-a4b-it-20260403::tools","model_id":"google/gemma-4-26b-a4b-it","canonical_id":"google/gemma-4-26b-a4b-it-20260403","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-04-06T06:45:01.382737+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 26B A4B ","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Gemma","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.35,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":26,"active_param_count_b":4,"is_moe":true,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-06T06:45:01.382737+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:10:18.669859+00:00","last_latency_ms":1638,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:29.686961+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::google/gemma-4-26b-a4b-it-20260403::low_temp","model_id":"google/gemma-4-26b-a4b-it","canonical_id":"google/gemma-4-26b-a4b-it-20260403","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-04-06T06:45:01.382737+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 26B A4B ","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Gemma","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.35,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":26,"active_param_count_b":4,"is_moe":true,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-06T06:45:01.382737+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":475,"first_unavailable_at":"2026-04-12T07:28:53.679506+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.483120+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-04-28::structured","model_id":"qwen/qwen3-30b-a3b","canonical_id":"qwen/qwen3-30b-a3b-04-28","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":40960,"max_completion_tokens":40960,"tokenizer":"Qwen3","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.28,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":4378,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":"qwen3-moe","decoder_type":"Sparse MoE","attention":"GQA with QK-Norm","arch_highlight":"High-capacity MoE design optimized for serving efficiency without a shared expert.","tech_report_url":"https://arxiv.org/pdf/2505.09388","hf_config_url":"https://huggingface.co/Qwen/Qwen3-235B-A22B/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.385556+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemma-4-26b-a4b-it-20260403::base","model_id":"google/gemma-4-26b-a4b-it","canonical_id":"google/gemma-4-26b-a4b-it-20260403","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-04-06T06:45:01.382737+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 26B A4B ","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Gemma","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.35,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":26,"active_param_count_b":4,"is_moe":true,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-06T06:45:01.382737+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":844,"first_unavailable_at":"2026-04-12T07:27:40.373624+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:05:50.944375+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-4-scout-17b-16e-instruct::structured","model_id":"meta-llama/llama-4-scout","canonical_id":"meta-llama/llama-4-scout-17b-16e-instruct","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 4 Scout","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":327680,"max_completion_tokens":16384,"tokenizer":"Llama4","pricing_input_per_1m":0.08,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":657,"first_unavailable_at":"2026-03-16T01:52:40.908000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.482428+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-32B::base","model_id":"Qwen/Qwen3-32B","canonical_id":"Qwen/Qwen3-32B","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-32B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.08,"pricing_output_per_1m":0.25,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":402,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.908411+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":4},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-instruct-2507::base","model_id":"qwen/qwen3-30b-a3b-instruct-2507","canonical_id":"qwen/qwen3-30b-a3b-instruct-2507","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Instruct 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen3","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:10:33.831979+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-instruct-2507::structured","model_id":"qwen/qwen3-30b-a3b-instruct-2507","canonical_id":"qwen/qwen3-30b-a3b-instruct-2507","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Instruct 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen3","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":2945,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.801856+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-30B-A3B::tools","model_id":"Qwen/Qwen3-30B-A3B","canonical_id":"Qwen/Qwen3-30B-A3B","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-30B-A3B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.09,"pricing_output_per_1m":0.45,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:10:18.669859+00:00","last_latency_ms":1048,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:54.496916+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-30B-A3B::low_temp","model_id":"Qwen/Qwen3-30B-A3B","canonical_id":"Qwen/Qwen3-30B-A3B","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-30B-A3B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.09,"pricing_output_per_1m":0.45,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":894,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.888872+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-30B-A3B::base","model_id":"Qwen/Qwen3-30B-A3B","canonical_id":"Qwen/Qwen3-30B-A3B","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-30B-A3B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.09,"pricing_output_per_1m":0.45,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":684,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.907659+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"openrouter::xiaomi/mimo-v2-flash-20251210::tools+reasoning","model_id":"xiaomi/mimo-v2-flash","canonical_id":"xiaomi/mimo-v2-flash-20251210","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Xiaomi: MiMo-V2-Flash","org":"xiaomi","org_name":"Xiaomi","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.29,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"MiMo series. Consumer-device focused training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1507,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.106560+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-instruct-2507::low_temp","model_id":"qwen/qwen3-30b-a3b-instruct-2507","canonical_id":"qwen/qwen3-30b-a3b-instruct-2507","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Instruct 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen3","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:32.057644+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-235B-A22B-Instruct-2507::structured","model_id":"Qwen/Qwen3-235B-A22B-Instruct-2507","canonical_id":"Qwen/Qwen3-235B-A22B-Instruct-2507","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-235B-A22B-Instruct-2507","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.09,"pricing_output_per_1m":0.58,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":364,"first_unavailable_at":"2026-04-12T13:14:11.279531+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.683382+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":5},{"variant_id":"huggingface::Qwen/Qwen3-235B-A22B-Instruct-2507::low_temp","model_id":"Qwen/Qwen3-235B-A22B-Instruct-2507","canonical_id":"Qwen/Qwen3-235B-A22B-Instruct-2507","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-235B-A22B-Instruct-2507","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.09,"pricing_output_per_1m":0.58,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":565,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.899921+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":5},{"variant_id":"huggingface::Qwen/Qwen3-235B-A22B-Instruct-2507::base","model_id":"Qwen/Qwen3-235B-A22B-Instruct-2507","canonical_id":"Qwen/Qwen3-235B-A22B-Instruct-2507","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-235B-A22B-Instruct-2507","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.09,"pricing_output_per_1m":0.58,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":331,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.980559+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":5},{"variant_id":"openrouter::xiaomi/mimo-v2-flash-20251210::base","model_id":"xiaomi/mimo-v2-flash","canonical_id":"xiaomi/mimo-v2-flash-20251210","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Xiaomi: MiMo-V2-Flash","org":"xiaomi","org_name":"Xiaomi","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.29,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"MiMo series. Consumer-device focused training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1744,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.184677+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::xiaomi/mimo-v2-flash-20251210::reasoning","model_id":"xiaomi/mimo-v2-flash","canonical_id":"xiaomi/mimo-v2-flash-20251210","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Xiaomi: MiMo-V2-Flash","org":"xiaomi","org_name":"Xiaomi","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.29,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"MiMo series. Consumer-device focused training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1137,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.500514+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-30b-a3b-instruct-2507::tools","model_id":"qwen/qwen3-30b-a3b-instruct-2507","canonical_id":"qwen/qwen3-30b-a3b-instruct-2507","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 30B A3B Instruct 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen3","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:52.015124+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::alibaba/tongyi-deepresearch-30b-a3b::tools","model_id":"alibaba/tongyi-deepresearch-30b-a3b","canonical_id":"alibaba/tongyi-deepresearch-30b-a3b","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tongyi DeepResearch 30B A3B","org":"alibaba","org_name":"Alibaba","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.44999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Alibaba cloud / Tongyi series.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-03-15T19:50:01.835000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:49.576104+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::alibaba/tongyi-deepresearch-30b-a3b::structured","model_id":"alibaba/tongyi-deepresearch-30b-a3b","canonical_id":"alibaba/tongyi-deepresearch-30b-a3b","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tongyi DeepResearch 30B A3B","org":"alibaba","org_name":"Alibaba","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.44999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Alibaba cloud / Tongyi series.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":239,"first_unavailable_at":"2026-03-15T19:50:01.835000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.119607+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::alibaba/tongyi-deepresearch-30b-a3b::reasoning","model_id":"alibaba/tongyi-deepresearch-30b-a3b","canonical_id":"alibaba/tongyi-deepresearch-30b-a3b","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tongyi DeepResearch 30B A3B","org":"alibaba","org_name":"Alibaba","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.44999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Alibaba cloud / Tongyi series.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1378,"first_unavailable_at":"2026-03-15T19:50:01.835000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.689387+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::alibaba/tongyi-deepresearch-30b-a3b::low_temp","model_id":"alibaba/tongyi-deepresearch-30b-a3b","canonical_id":"alibaba/tongyi-deepresearch-30b-a3b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tongyi DeepResearch 30B A3B","org":"alibaba","org_name":"Alibaba","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.44999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Alibaba cloud / Tongyi series.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:36.850326+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::alibaba/tongyi-deepresearch-30b-a3b::base","model_id":"alibaba/tongyi-deepresearch-30b-a3b","canonical_id":"alibaba/tongyi-deepresearch-30b-a3b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tongyi DeepResearch 30B A3B","org":"alibaba","org_name":"Alibaba","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09,"pricing_output_per_1m":0.44999999999999996,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":30,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Alibaba cloud / Tongyi series.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:30:52.062026+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-next-80b-a3b-thinking-2509::reasoning","model_id":"qwen/qwen3-next-80b-a3b-thinking","canonical_id":"qwen/qwen3-next-80b-a3b-thinking-2509","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Next 80B A3B Thinking","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":32768,"tokenizer":"Qwen3","pricing_input_per_1m":0.0975,"pricing_output_per_1m":0.78,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1566,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.691128+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-next-80b-a3b-thinking-2509::structured","model_id":"qwen/qwen3-next-80b-a3b-thinking","canonical_id":"qwen/qwen3-next-80b-a3b-thinking-2509","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Next 80B A3B Thinking","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":32768,"tokenizer":"Qwen3","pricing_input_per_1m":0.0975,"pricing_output_per_1m":0.78,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1173,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.183916+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-next-80b-a3b-thinking-2509::tools+reasoning","model_id":"qwen/qwen3-next-80b-a3b-thinking","canonical_id":"qwen/qwen3-next-80b-a3b-thinking-2509","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Next 80B A3B Thinking","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":32768,"tokenizer":"Qwen3","pricing_input_per_1m":0.0975,"pricing_output_per_1m":0.78,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1994,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.303334+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite::tools","model_id":"google/gemini-2.5-flash-lite","canonical_id":"google/gemini-2.5-flash-lite","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:01.884098+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::mistralai/ministral-3b-2512::structured","model_id":"mistralai/ministral-3b-2512","canonical_id":"mistralai/ministral-3b-2512","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 3B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1188,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.983509+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::nvidia/llama-3.3-nemotron-super-49b-v1.5::base","model_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","canonical_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Llama 3.3 Nemotron Super 49B V1.5","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":49,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:58:53.080993+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite::structured","model_id":"google/gemini-2.5-flash-lite","canonical_id":"google/gemini-2.5-flash-lite","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":333,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.226221+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::mistralai/ministral-3b-2512::tools","model_id":"mistralai/ministral-3b-2512","canonical_id":"mistralai/ministral-3b-2512","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 3B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:03.006821+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::mistralai/ministral-3b-2512::low_temp","model_id":"mistralai/ministral-3b-2512","canonical_id":"mistralai/ministral-3b-2512","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 3B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":298,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.986827+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/ministral-3b-2512::base","model_id":"mistralai/ministral-3b-2512","canonical_id":"mistralai/ministral-3b-2512","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 3B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.09999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":3,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":541,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.281864+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/llama-3.3-nemotron-super-49b-v1.5::low_temp","model_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","canonical_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Llama 3.3 Nemotron Super 49B V1.5","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":49,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:47.930424+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite-preview-09-2025::tools+reasoning","model_id":"google/gemini-2.5-flash-lite-preview-09-2025","canonical_id":"google/gemini-2.5-flash-lite-preview-09-2025","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite Preview 09-2025","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1319,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.292243+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite-preview-09-2025::low_temp","model_id":"google/gemini-2.5-flash-lite-preview-09-2025","canonical_id":"google/gemini-2.5-flash-lite-preview-09-2025","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite Preview 09-2025","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:45.272552+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite::low_temp","model_id":"google/gemini-2.5-flash-lite","canonical_id":"google/gemini-2.5-flash-lite","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:42.794017+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite::base","model_id":"google/gemini-2.5-flash-lite","canonical_id":"google/gemini-2.5-flash-lite","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:10:37.023584+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.0-flash-001::tools","model_id":"google/gemini-2.0-flash-001","canonical_id":"google/gemini-2.0-flash-001","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.0 Flash","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":"2026-06-01","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:57.757074+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.0-flash-001::structured","model_id":"google/gemini-2.0-flash-001","canonical_id":"google/gemini-2.0-flash-001","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.0 Flash","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":"2026-06-01","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":420,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.500200+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.0-flash-001::base","model_id":"google/gemini-2.0-flash-001","canonical_id":"google/gemini-2.0-flash-001","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.0 Flash","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":"2026-06-01","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:31:07.820025+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nvidia/llama-3.3-nemotron-super-49b-v1.5::tools","model_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","canonical_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Llama 3.3 Nemotron Super 49B V1.5","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":49,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:56.552152+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::rekaai/reka-flash-3::low_temp","model_id":"rekaai/reka-flash-3","canonical_id":"rekaai/reka-flash-3","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Reka Flash 3","org":"rekaai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":65536,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:39.220211+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","seed","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::rekaai/reka-flash-3::base","model_id":"rekaai/reka-flash-3","canonical_id":"rekaai/reka-flash-3","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Reka Flash 3","org":"rekaai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":65536,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:58:47.275117+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","seed","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::rekaai/reka-flash-3::reasoning","model_id":"rekaai/reka-flash-3","canonical_id":"rekaai/reka-flash-3","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Reka Flash 3","org":"rekaai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":65536,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2041,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.999299+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","seed","stop","temperature","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-9b-20260310::tools+reasoning","model_id":"qwen/qwen3.5-9b","canonical_id":"qwen/qwen3.5-9b-20260310","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-9B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":2481,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:04.954582+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::mistralai/voxtral-small-24b-2507::low_temp","model_id":"mistralai/voxtral-small-24b-2507","canonical_id":"mistralai/voxtral-small-24b-2507","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Voxtral Small 24B 2507","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32000,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":24,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","audio"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:40.393890+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/voxtral-small-24b-2507::tools","model_id":"mistralai/voxtral-small-24b-2507","canonical_id":"mistralai/voxtral-small-24b-2507","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Voxtral Small 24B 2507","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32000,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":24,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","audio"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:00.663691+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::nvidia/llama-3.3-nemotron-super-49b-v1.5::tools+reasoning","model_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","canonical_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Llama 3.3 Nemotron Super 49B V1.5","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":49,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1719,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.210530+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::mistralai/voxtral-small-24b-2507::base","model_id":"mistralai/voxtral-small-24b-2507","canonical_id":"mistralai/voxtral-small-24b-2507","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Voxtral Small 24B 2507","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32000,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":24,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","audio"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:31:09.168364+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite::reasoning","model_id":"google/gemini-2.5-flash-lite","canonical_id":"google/gemini-2.5-flash-lite","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1294,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.789857+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-creative-20251216::base","model_id":"mistralai/mistral-small-creative","canonical_id":"mistralai/mistral-small-creative-20251216","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small Creative","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":"2026-04-30","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":548,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.183367+00:00","status_source":"live_traffic","claimed_capabilities":["tool_choice","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/voxtral-small-24b-2507::structured","model_id":"mistralai/voxtral-small-24b-2507","canonical_id":"mistralai/voxtral-small-24b-2507","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Voxtral Small 24B 2507","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32000,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":24,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","audio"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":777,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.797983+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-9b-20260310::low_temp","model_id":"qwen/qwen3.5-9b","canonical_id":"qwen/qwen3.5-9b-20260310","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-9B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1069,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.789414+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/devstral-small-2507::structured","model_id":"mistralai/devstral-small","canonical_id":"mistralai/devstral-small-2507","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Devstral Small 1.1","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":413,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.285500+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-9b-20260310::base","model_id":"qwen/qwen3.5-9b","canonical_id":"qwen/qwen3.5-9b-20260310","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-9B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":593,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.918270+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::bytedance/ui-tars-1.5-7b::low_temp","model_id":"bytedance/ui-tars-1.5-7b","canonical_id":"bytedance/ui-tars-1.5-7b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance: UI-TARS 7B ","org":"bytedance","org_name":"ByteDance","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":2048,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"ByteDance product models.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:43.946796+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.0-flash-001::low_temp","model_id":"google/gemini-2.0-flash-001","canonical_id":"google/gemini-2.0-flash-001","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.0 Flash","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":8192,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":"2026-06-01","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:44:46.411281+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-2.0-mini-20260224::structured","model_id":"bytedance-seed/seed-2.0-mini","canonical_id":"bytedance-seed/seed-2.0-mini-20260224","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed-2.0-Mini","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":4058,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.791820+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-2.0-mini-20260224::reasoning","model_id":"bytedance-seed/seed-2.0-mini","canonical_id":"bytedance-seed/seed-2.0-mini-20260224","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed-2.0-Mini","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2761,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.396146+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-2.0-mini-20260224::low_temp","model_id":"bytedance-seed/seed-2.0-mini","canonical_id":"bytedance-seed/seed-2.0-mini-20260224","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed-2.0-Mini","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":553,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.797282+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::bytedance-seed/seed-2.0-mini-20260224::base","model_id":"bytedance-seed/seed-2.0-mini","canonical_id":"bytedance-seed/seed-2.0-mini-20260224","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance Seed: Seed-2.0-Mini","org":"bytedance-seed","org_name":"ByteDance (Seed Team)","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Seed series. ByteDance's frontier model research. Access to TikTok/Douyin data. Strong multimodal potential.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":961,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.982510+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","reasoning","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-9b-20260310::reasoning","model_id":"qwen/qwen3.5-9b","canonical_id":"qwen/qwen3.5-9b-20260310","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-9B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1731,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.386289+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::bytedance/ui-tars-1.5-7b::base","model_id":"bytedance/ui-tars-1.5-7b","canonical_id":"bytedance/ui-tars-1.5-7b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"ByteDance: UI-TARS 7B ","org":"bytedance","org_name":"ByteDance","country":"China","city":"Beijing","org_type":"big-tech","open_weights":null,"context_length":128000,"max_completion_tokens":2048,"tokenizer":"Other","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"ByteDance product models.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:58:50.767404+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","presence_penalty","repetition_penalty","seed","stop","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite-preview-09-2025::structured","model_id":"google/gemini-2.5-flash-lite-preview-09-2025","canonical_id":"google/gemini-2.5-flash-lite-preview-09-2025","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite Preview 09-2025","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":435,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.111317+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite-preview-09-2025::reasoning","model_id":"google/gemini-2.5-flash-lite-preview-09-2025","canonical_id":"google/gemini-2.5-flash-lite-preview-09-2025","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite Preview 09-2025","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1663,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.680705+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-creative-20251216::low_temp","model_id":"mistralai/mistral-small-creative","canonical_id":"mistralai/mistral-small-creative-20251216","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small Creative","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.3,"tags":"text_generation","release_date":null,"expiration_date":"2026-04-30","param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1120,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.906734+00:00","status_source":"live_traffic","claimed_capabilities":["tool_choice","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite-preview-09-2025::tools","model_id":"google/gemini-2.5-flash-lite-preview-09-2025","canonical_id":"google/gemini-2.5-flash-lite-preview-09-2025","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite Preview 09-2025","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:37:59.180926+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::nvidia/llama-3.3-nemotron-super-49b-v1.5::reasoning","model_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","canonical_id":"nvidia/llama-3.3-nemotron-super-49b-v1.5","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NVIDIA: Llama 3.3 Nemotron Super 49B V1.5","org":"nvidia","org_name":"NVIDIA","country":"USA","city":"Santa Clara, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":49,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Nemotron series. Built on top of Llama. Optimized for inference on NVIDIA hardware. Strong technical/scientific domains.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1651,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.609377+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite::tools+reasoning","model_id":"google/gemini-2.5-flash-lite","canonical_id":"google/gemini-2.5-flash-lite","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":985,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.408131+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-9b-20260310::structured","model_id":"qwen/qwen3.5-9b","canonical_id":"qwen/qwen3.5-9b-20260310","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-15T23:31:51.523000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-9B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Qwen3","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":9,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":815,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.720246+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemini-2.5-flash-lite-preview-09-2025::base","model_id":"google/gemini-2.5-flash-lite-preview-09-2025","canonical_id":"google/gemini-2.5-flash-lite-preview-09-2025","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemini 2.5 Flash Lite Preview 09-2025","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":65535,"tokenizer":"Gemini","pricing_input_per_1m":0.09999999999999999,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file","audio","video"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:11:38.117477+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::utter-project/EuroLLM-22B-Instruct-2512::low_temp","model_id":"utter-project/EuroLLM-22B-Instruct-2512","canonical_id":"utter-project/EuroLLM-22B-Instruct-2512","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-06T06:45:19.840772+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"EuroLLM-22B-Instruct-2512","org":"utter-project","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":null,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.1,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":22,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-06T06:45:19.840772+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1464,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.908683+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::swiss-ai/Apertus-8B-Instruct-2509::structured","model_id":"swiss-ai/Apertus-8B-Instruct-2509","canonical_id":"swiss-ai/Apertus-8B-Instruct-2509","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Apertus-8B-Instruct-2509","org":"swiss-ai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":null,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.1,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1389,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.598878+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":1},{"variant_id":"huggingface::swiss-ai/Apertus-8B-Instruct-2509::low_temp","model_id":"swiss-ai/Apertus-8B-Instruct-2509","canonical_id":"swiss-ai/Apertus-8B-Instruct-2509","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Apertus-8B-Instruct-2509","org":"swiss-ai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":null,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.1,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1906,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.690140+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::swiss-ai/Apertus-8B-Instruct-2509::base","model_id":"swiss-ai/Apertus-8B-Instruct-2509","canonical_id":"swiss-ai/Apertus-8B-Instruct-2509","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Apertus-8B-Instruct-2509","org":"swiss-ai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":null,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.1,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1400,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.708962+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B::low_temp","model_id":"deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B","canonical_id":"deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek-R1-Distill-Qwen-1.5B","org":"deepseek-ai","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.1,"pricing_output_per_1m":0.1,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":1.5,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":675,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.508285+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B::base","model_id":"deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B","canonical_id":"deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek-R1-Distill-Qwen-1.5B","org":"deepseek-ai","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.1,"pricing_output_per_1m":0.1,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":1.5,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1063,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.590803+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::allenai/Olmo-3-7B-Instruct::low_temp","model_id":"allenai/Olmo-3-7B-Instruct","canonical_id":"allenai/Olmo-3-7B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-06T06:45:19.840772+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Olmo-3-7B-Instruct","org":"allenai","org_name":"Allen Institute for AI","country":"USA","city":"Seattle, WA","org_type":"non-profit","open_weights":true,"context_length":null,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.1,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"OLMo series. Fully open training data (Dolma). Strong research reproducibility focus. Non-profit academic lab.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-06T06:45:19.840772+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1749,"first_unavailable_at":"2026-04-12T13:13:20.285377+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.907547+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::allenai/Olmo-3-7B-Instruct::base","model_id":"allenai/Olmo-3-7B-Instruct","canonical_id":"allenai/Olmo-3-7B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-06T06:45:19.840772+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Olmo-3-7B-Instruct","org":"allenai","org_name":"Allen Institute for AI","country":"USA","city":"Seattle, WA","org_type":"non-profit","open_weights":true,"context_length":null,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.1,"pricing_output_per_1m":0.2,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"OLMo series. Fully open training data (Dolma). Strong research reproducibility focus. Non-profit academic lab.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-06T06:45:19.840772+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1405,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.991200+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"openrouter::mistralai/mistral-7b-instruct-v0.1::base","model_id":"mistralai/mistral-7b-instruct-v0.1","canonical_id":"mistralai/mistral-7b-instruct-v0.1","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral 7B Instruct v0.1","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":2824,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.11,"pricing_output_per_1m":0.19,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:12:23.401770+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/mistral-7b-instruct-v0.1::low_temp","model_id":"mistralai/mistral-7b-instruct-v0.1","canonical_id":"mistralai/mistral-7b-instruct-v0.1","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral 7B Instruct v0.1","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":2824,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.11,"pricing_output_per_1m":0.19,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:10.007477+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","seed","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-m2.5-20260211::structured","model_id":"minimax/minimax-m2.5","canonical_id":"minimax/minimax-m2.5-20260211","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax M2.5","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":196608,"max_completion_tokens":65536,"tokenizer":"Other","pricing_input_per_1m":0.118,"pricing_output_per_1m":0.9900000000000001,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","parallel_tool_calls","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":9080,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.811261+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","parallel_tool_calls","presence_penalty","reasoning","reasoning_effort","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen-2.5-72b-instruct::structured","model_id":"qwen/qwen-2.5-72b-instruct","canonical_id":"qwen/qwen-2.5-72b-instruct","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5 72B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":16384,"tokenizer":"Qwen","pricing_input_per_1m":0.12,"pricing_output_per_1m":0.39,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":72,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1791,"first_unavailable_at":"2026-03-15T19:50:01.835000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.881353+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen-2.5-72b-instruct::tools","model_id":"qwen/qwen-2.5-72b-instruct","canonical_id":"qwen/qwen-2.5-72b-instruct","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5 72B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":16384,"tokenizer":"Qwen","pricing_input_per_1m":0.12,"pricing_output_per_1m":0.39,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":72,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:07.245541+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::qwen/qwen-2.5-72b-instruct::base","model_id":"qwen/qwen-2.5-72b-instruct","canonical_id":"qwen/qwen-2.5-72b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5 72B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":16384,"tokenizer":"Qwen","pricing_input_per_1m":0.12,"pricing_output_per_1m":0.39,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":72,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:59:26.113559+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen-2.5-72b-instruct::low_temp","model_id":"qwen/qwen-2.5-72b-instruct","canonical_id":"qwen/qwen-2.5-72b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen2.5 72B Instruct","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":32768,"max_completion_tokens":16384,"tokenizer":"Qwen","pricing_input_per_1m":0.12,"pricing_output_per_1m":0.39,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":72,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:11.338389+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-3.3-70b-instruct::structured","model_id":"meta-llama/llama-3.3-70b-instruct","canonical_id":"meta-llama/llama-3.3-70b-instruct","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 3.3 70B Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Llama3","pricing_input_per_1m":0.12,"pricing_output_per_1m":0.38,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":179,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.513273+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::zai-org/GLM-4.5-Air::tools","model_id":"zai-org/GLM-4.5-Air","canonical_id":"zai-org/GLM-4.5-Air","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"GLM-4.5-Air","org":"zai-org","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.13,"pricing_output_per_1m":0.85,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:10:18.669859+00:00","last_latency_ms":815,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:15.127105+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":2},{"variant_id":"huggingface::zai-org/GLM-4.5-Air::low_temp","model_id":"zai-org/GLM-4.5-Air","canonical_id":"zai-org/GLM-4.5-Air","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"GLM-4.5-Air","org":"zai-org","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.13,"pricing_output_per_1m":0.85,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1582,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.686530+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"huggingface::zai-org/GLM-4.5-Air::base","model_id":"zai-org/GLM-4.5-Air","canonical_id":"zai-org/GLM-4.5-Air","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"GLM-4.5-Air","org":"zai-org","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.13,"pricing_output_per_1m":0.85,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T07:59:31.076769+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"openrouter::qwen/qwen3-235b-a22b-thinking-2507::base","model_id":"qwen/qwen3-235b-a22b-thinking-2507","canonical_id":"qwen/qwen3-235b-a22b-thinking-2507","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 235B A22B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen3","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:31:46.168448+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-235b-a22b-thinking-2507::low_temp","model_id":"qwen/qwen3-235b-a22b-thinking-2507","canonical_id":"qwen/qwen3-235b-a22b-thinking-2507","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 235B A22B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen3","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:28.613759+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-235b-a22b-thinking-2507::reasoning","model_id":"qwen/qwen3-235b-a22b-thinking-2507","canonical_id":"qwen/qwen3-235b-a22b-thinking-2507","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 235B A22B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen3","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":2227,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.788027+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-235b-a22b-thinking-2507::tools","model_id":"qwen/qwen3-235b-a22b-thinking-2507","canonical_id":"qwen/qwen3-235b-a22b-thinking-2507","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 235B A22B Thinking 2507","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen3","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:33.509971+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::z-ai/glm-4.5-air::structured","model_id":"z-ai/glm-4.5-air","canonical_id":"z-ai/glm-4.5-air","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Z.ai: GLM 4.5 Air","org":"z-ai","org_name":"Zhipu AI","country":"China","city":"Beijing","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":98304,"tokenizer":"Other","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.85,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","seed","stop","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GLM series. Spin-off from Tsinghua University. Strong academic and Chinese language. Bilingual architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":565,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.222491+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::google/gemma-4-31b-it-20260402::tools","model_id":"google/gemma-4-31b-it","canonical_id":"google/gemma-4-31b-it-20260402","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-04-06T06:45:01.382737+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 31B","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Gemma","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.38,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":31,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-03T04:27:56.857119+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:10:18.669859+00:00","last_latency_ms":1075,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T23:19:22.773986+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-4-70b::reasoning","model_id":"nousresearch/hermes-4-70b","canonical_id":"nousresearch/hermes-4-70b","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nous: Hermes 4 70B","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1520,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.700529+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","temperature","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::google/gemma-4-31b-it-20260402::base","model_id":"google/gemma-4-31b-it","canonical_id":"google/gemma-4-31b-it-20260402","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-04-03T04:27:56.857119+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Google: Gemma 4 31B","org":"google","org_name":"Google DeepMind","country":"USA","city":"Mountain View, CA / London, UK","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Gemma","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.38,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":31,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["image","text","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Gemini series + Gemma open models. Strong multilingual, scientific, and multimodal training.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-03T04:27:56.857119+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":371,"first_unavailable_at":"2026-04-12T01:03:07.980898+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-13T21:06:03.533827+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-4-70b::base","model_id":"nousresearch/hermes-4-70b","canonical_id":"nousresearch/hermes-4-70b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nous: Hermes 4 70B","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:09:04.126932+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-4-70b::low_temp","model_id":"nousresearch/hermes-4-70b","canonical_id":"nousresearch/hermes-4-70b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nous: Hermes 4 70B","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":131072,"max_completion_tokens":null,"tokenizer":"Llama3","pricing_input_per_1m":0.13,"pricing_output_per_1m":0.39999999999999997,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:14.920804+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nex-agi/deepseek-v3.1-nex-n1::structured","model_id":"nex-agi/deepseek-v3.1-nex-n1","canonical_id":"nex-agi/deepseek-v3.1-nex-n1","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nex AGI: DeepSeek V3.1 Nex N1","org":"nex-agi","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":131072,"max_completion_tokens":163840,"tokenizer":"DeepSeek","pricing_input_per_1m":0.135,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":973,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.911647+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::meta-llama/Llama-3.3-70B-Instruct::search","model_id":"meta-llama/Llama-3.3-70B-Instruct","canonical_id":"meta-llama/Llama-3.3-70B-Instruct","config_key":"search","variant_key":"search","interface":"huggingface","first_seen_at":"2026-04-13T19:43:59.804778+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Llama-3.3-70B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.135,"pricing_output_per_1m":0.4,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":193,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.308136+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::meta-llama/Llama-3.3-70B-Instruct::low_temp","model_id":"meta-llama/Llama-3.3-70B-Instruct","canonical_id":"meta-llama/Llama-3.3-70B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Llama-3.3-70B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.135,"pricing_output_per_1m":0.4,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":275,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.488450+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":8},{"variant_id":"huggingface::meta-llama/Llama-3.3-70B-Instruct::base","model_id":"meta-llama/Llama-3.3-70B-Instruct","canonical_id":"meta-llama/Llama-3.3-70B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Llama-3.3-70B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.135,"pricing_output_per_1m":0.4,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":168,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.510012+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":8},{"variant_id":"huggingface::meta-llama/Llama-3.3-70B-Instruct::tools","model_id":"meta-llama/Llama-3.3-70B-Instruct","canonical_id":"meta-llama/Llama-3.3-70B-Instruct","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Llama-3.3-70B-Instruct","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.135,"pricing_output_per_1m":0.4,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":70,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:10:18.669859+00:00","last_latency_ms":404,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T10:25:25.779742+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":8},{"variant_id":"openrouter::nex-agi/deepseek-v3.1-nex-n1::base","model_id":"nex-agi/deepseek-v3.1-nex-n1","canonical_id":"nex-agi/deepseek-v3.1-nex-n1","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nex AGI: DeepSeek V3.1 Nex N1","org":"nex-agi","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":131072,"max_completion_tokens":163840,"tokenizer":"DeepSeek","pricing_input_per_1m":0.135,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1054,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.215797+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nex-agi/deepseek-v3.1-nex-n1::low_temp","model_id":"nex-agi/deepseek-v3.1-nex-n1","canonical_id":"nex-agi/deepseek-v3.1-nex-n1","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Nex AGI: DeepSeek V3.1 Nex N1","org":"nex-agi","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":131072,"max_completion_tokens":163840,"tokenizer":"DeepSeek","pricing_input_per_1m":0.135,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1057,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.979501+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-2-pro-llama-3-8b::structured","model_id":"nousresearch/hermes-2-pro-llama-3-8b","canonical_id":"nousresearch/hermes-2-pro-llama-3-8b","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NousResearch: Hermes 2 Pro - Llama-3 8B","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":8192,"max_completion_tokens":8192,"tokenizer":"Llama3","pricing_input_per_1m":0.14,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":3334,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.684331+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-2-pro-llama-3-8b::low_temp","model_id":"nousresearch/hermes-2-pro-llama-3-8b","canonical_id":"nousresearch/hermes-2-pro-llama-3-8b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NousResearch: Hermes 2 Pro - Llama-3 8B","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":8192,"max_completion_tokens":8192,"tokenizer":"Llama3","pricing_input_per_1m":0.14,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:19.848280+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::tencent/hunyuan-a13b-instruct::base","model_id":"tencent/hunyuan-a13b-instruct","canonical_id":"tencent/hunyuan-a13b-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tencent: Hunyuan A13B Instruct","org":"tencent","org_name":"Tencent","country":"China","city":"Shenzhen","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.14,"pricing_output_per_1m":0.5700000000000001,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":13,"active_param_count_b":13,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","reasoning","response_format","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hunyuan series. Access to WeChat/QQ messaging data. Strong Chinese social and conversational language.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:31:37.434638+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","reasoning","response_format","structured_outputs","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::tencent/hunyuan-a13b-instruct::low_temp","model_id":"tencent/hunyuan-a13b-instruct","canonical_id":"tencent/hunyuan-a13b-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tencent: Hunyuan A13B Instruct","org":"tencent","org_name":"Tencent","country":"China","city":"Shenzhen","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.14,"pricing_output_per_1m":0.5700000000000001,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":13,"active_param_count_b":13,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","reasoning","response_format","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hunyuan series. Access to WeChat/QQ messaging data. Strong Chinese social and conversational language.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:22.612841+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","reasoning","response_format","structured_outputs","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::tencent/hunyuan-a13b-instruct::reasoning","model_id":"tencent/hunyuan-a13b-instruct","canonical_id":"tencent/hunyuan-a13b-instruct","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tencent: Hunyuan A13B Instruct","org":"tencent","org_name":"Tencent","country":"China","city":"Shenzhen","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.14,"pricing_output_per_1m":0.5700000000000001,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":13,"active_param_count_b":13,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","reasoning","response_format","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hunyuan series. Access to WeChat/QQ messaging data. Strong Chinese social and conversational language.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":3419,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.796837+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","reasoning","response_format","structured_outputs","temperature","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::tencent/hunyuan-a13b-instruct::structured","model_id":"tencent/hunyuan-a13b-instruct","canonical_id":"tencent/hunyuan-a13b-instruct","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Tencent: Hunyuan A13B Instruct","org":"tencent","org_name":"Tencent","country":"China","city":"Shenzhen","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.14,"pricing_output_per_1m":0.5700000000000001,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":13,"active_param_count_b":13,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","reasoning","response_format","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hunyuan series. Access to WeChat/QQ messaging data. Strong Chinese social and conversational language.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1533,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.290619+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","reasoning","response_format","structured_outputs","temperature","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::nousresearch/hermes-2-pro-llama-3-8b::base","model_id":"nousresearch/hermes-2-pro-llama-3-8b","canonical_id":"nousresearch/hermes-2-pro-llama-3-8b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"NousResearch: Hermes 2 Pro - Llama-3 8B","org":"nousresearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":null,"context_length":8192,"max_completion_tokens":8192,"tokenizer":"Llama3","pricing_input_per_1m":0.14,"pricing_output_per_1m":0.14,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:31:34.382364+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::NousResearch/Hermes-2-Pro-Llama-3-8B::low_temp","model_id":"NousResearch/Hermes-2-Pro-Llama-3-8B","canonical_id":"NousResearch/Hermes-2-Pro-Llama-3-8B","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Hermes-2-Pro-Llama-3-8B","org":"NousResearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.14,"pricing_output_per_1m":0.14,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:24.460193+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::NousResearch/Hermes-2-Pro-Llama-3-8B::base","model_id":"NousResearch/Hermes-2-Pro-Llama-3-8B","canonical_id":"NousResearch/Hermes-2-Pro-Llama-3-8B","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Hermes-2-Pro-Llama-3-8B","org":"NousResearch","org_name":"NousResearch","country":"USA","city":"Distributed","org_type":"community","open_weights":true,"context_length":8192,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.14,"pricing_output_per_1m":0.14,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Hermes series. Leading community fine-tune lab for Llama models.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:15:26.794532+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"openrouter::openai/gpt-4o-mini-2024-07-18::base","model_id":"openai/gpt-4o-mini-2024-07-18","canonical_id":"openai/gpt-4o-mini-2024-07-18","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini (2024-07-18)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:31:53.030877+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwq-32b::low_temp","model_id":"qwen/qwq-32b","canonical_id":"qwen/qwq-32b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: QwQ 32B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.58,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","reasoning","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:44.522321+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","reasoning","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-4o-mini-2024-07-18::structured","model_id":"openai/gpt-4o-mini-2024-07-18","canonical_id":"openai/gpt-4o-mini-2024-07-18","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini (2024-07-18)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":684,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.603548+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-2603::reasoning","model_id":"mistralai/mistral-small-2603","canonical_id":"mistralai/mistral-small-2603","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 4","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1436,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.995988+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::openai/gpt-4o-mini-2024-07-18::low_temp","model_id":"openai/gpt-4o-mini-2024-07-18","canonical_id":"openai/gpt-4o-mini-2024-07-18","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini (2024-07-18)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:46:26.954940+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::deepseek-ai/DeepSeek-R1-Distill-Qwen-7B::low_temp","model_id":"deepseek-ai/DeepSeek-R1-Distill-Qwen-7B","canonical_id":"deepseek-ai/DeepSeek-R1-Distill-Qwen-7B","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek-R1-Distill-Qwen-7B","org":"deepseek-ai","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1009,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.509486+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"openrouter::qwen/qwq-32b::tools","model_id":"qwen/qwq-32b","canonical_id":"qwen/qwq-32b","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: QwQ 32B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.58,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","reasoning","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":"2026-03-15T19:50:01.835000+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:45.101532+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","reasoning","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::mistralai/ministral-8b-2512::low_temp","model_id":"mistralai/ministral-8b-2512","canonical_id":"mistralai/ministral-8b-2512","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 8B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":696,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.985774+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/ministral-8b-2512::structured","model_id":"mistralai/ministral-8b-2512","canonical_id":"mistralai/ministral-8b-2512","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 8B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":366,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.918862+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::mistralai/ministral-8b-2512::base","model_id":"mistralai/ministral-8b-2512","canonical_id":"mistralai/ministral-8b-2512","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 8B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":292,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.279706+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/ministral-8b-2512::tools","model_id":"mistralai/ministral-8b-2512","canonical_id":"mistralai/ministral-8b-2512","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 8B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":8,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:34.719399+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-2603::low_temp","model_id":"mistralai/mistral-small-2603","canonical_id":"mistralai/mistral-small-2603","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 4","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":404,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.783351+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::essentialai/rnj-1-instruct::tools","model_id":"essentialai/rnj-1-instruct","canonical_id":"essentialai/rnj-1-instruct","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"EssentialAI: Rnj 1 Instruct","org":"essentialai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:20.018628+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::essentialai/rnj-1-instruct::structured","model_id":"essentialai/rnj-1-instruct","canonical_id":"essentialai/rnj-1-instruct","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"EssentialAI: Rnj 1 Instruct","org":"essentialai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":220,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.913446+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::essentialai/rnj-1-instruct::low_temp","model_id":"essentialai/rnj-1-instruct","canonical_id":"essentialai/rnj-1-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"EssentialAI: Rnj 1 Instruct","org":"essentialai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":270,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.981032+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::essentialai/rnj-1-instruct::base","model_id":"essentialai/rnj-1-instruct","canonical_id":"essentialai/rnj-1-instruct","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"EssentialAI: Rnj 1 Instruct","org":"essentialai","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":282,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.217481+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-Next-80B-A3B-Thinking::tools","model_id":"Qwen/Qwen3-Next-80B-A3B-Thinking","canonical_id":"Qwen/Qwen3-Next-80B-A3B-Thinking","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Next-80B-A3B-Thinking","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:54.812104+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-Next-80B-A3B-Thinking::low_temp","model_id":"Qwen/Qwen3-Next-80B-A3B-Thinking","canonical_id":"Qwen/Qwen3-Next-80B-A3B-Thinking","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Next-80B-A3B-Thinking","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:03:43.384368+00:00","last_latency_ms":3966,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:48.359406+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-Next-80B-A3B-Thinking::base","model_id":"Qwen/Qwen3-Next-80B-A3B-Thinking","canonical_id":"Qwen/Qwen3-Next-80B-A3B-Thinking","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Next-80B-A3B-Thinking","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:03:07.980898+00:00","last_latency_ms":6002,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:00:10.742524+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"openrouter::qwen/qwen3-coder-next-2025-02-03::structured","model_id":"qwen/qwen3-coder-next","canonical_id":"qwen/qwen3-coder-next-2025-02-03","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder Next","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.7999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1211,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.817955+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-Next-80B-A3B-Instruct::tools","model_id":"Qwen/Qwen3-Next-80B-A3B-Instruct","canonical_id":"Qwen/Qwen3-Next-80B-A3B-Instruct","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Next-80B-A3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:10:18.669859+00:00","last_latency_ms":1121,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:57.228188+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":2},{"variant_id":"huggingface::Qwen/Qwen3-Next-80B-A3B-Instruct::low_temp","model_id":"Qwen/Qwen3-Next-80B-A3B-Instruct","canonical_id":"Qwen/Qwen3-Next-80B-A3B-Instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Next-80B-A3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":955,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.905379+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"huggingface::Qwen/Qwen3-Next-80B-A3B-Instruct::base","model_id":"Qwen/Qwen3-Next-80B-A3B-Instruct","canonical_id":"Qwen/Qwen3-Next-80B-A3B-Instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Next-80B-A3B-Instruct","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":80,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":997,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.988807+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":2},{"variant_id":"openrouter::openai/gpt-4o-mini::base","model_id":"openai/gpt-4o-mini","canonical_id":"openai/gpt-4o-mini","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_completion_tokens","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:09:22.529072+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_completion_tokens","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-2603::base","model_id":"mistralai/mistral-small-2603","canonical_id":"mistralai/mistral-small-2603","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 4","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":423,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.910275+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::meta-llama/llama-4-maverick-17b-128e-instruct::structured","model_id":"meta-llama/llama-4-maverick","canonical_id":"meta-llama/llama-4-maverick-17b-128e-instruct","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Meta: Llama 4 Maverick","org":"meta-llama","org_name":"Meta AI","country":"USA","city":"Menlo Park, CA","org_type":"big-tech","open_weights":null,"context_length":1048576,"max_completion_tokens":16384,"tokenizer":"Llama4","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Llama series. Open weights. Heavy Common Crawl + open web. Strong English, decent multilingual from Llama 3 onwards.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":625,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":"llama-4-maverick","decoder_type":"Sparse MoE","attention":"GQA","arch_highlight":"Alternates dense and MoE blocks and uses fewer, larger experts than DeepSeek V3.","tech_report_url":"https://ai.meta.com/blog/llama-4-multimodal-intelligence/","hf_config_url":"https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.409133+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-2603::structured","model_id":"mistralai/mistral-small-2603","canonical_id":"mistralai/mistral-small-2603","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 4","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":840,"first_unavailable_at":"2026-04-12T13:14:11.279531+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.714911+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-2603::tools+reasoning","model_id":"mistralai/mistral-small-2603","canonical_id":"mistralai/mistral-small-2603","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 4","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1652,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.711063+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-small-2603::tools","model_id":"mistralai/mistral-small-2603","canonical_id":"mistralai/mistral-small-2603","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Mistral Small 4","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:11:32.167307+00:00","last_latency_ms":505,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:38:58.870411+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3.1::tools+reasoning","model_id":"deepseek/deepseek-chat-v3.1","canonical_id":"deepseek/deepseek-chat-v3.1","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3.1","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":7168,"tokenizer":"DeepSeek","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.75,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":15989,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.314962+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3.1::tools","model_id":"deepseek/deepseek-chat-v3.1","canonical_id":"deepseek/deepseek-chat-v3.1","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3.1","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":7168,"tokenizer":"DeepSeek","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.75,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:03.760710+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3.1::structured","model_id":"deepseek/deepseek-chat-v3.1","canonical_id":"deepseek/deepseek-chat-v3.1","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3.1","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":7168,"tokenizer":"DeepSeek","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.75,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1689,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.198200+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3.1::reasoning","model_id":"deepseek/deepseek-chat-v3.1","canonical_id":"deepseek/deepseek-chat-v3.1","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3.1","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":7168,"tokenizer":"DeepSeek","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.75,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":8542,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.704044+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3.1::low_temp","model_id":"deepseek/deepseek-chat-v3.1","canonical_id":"deepseek/deepseek-chat-v3.1","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3.1","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":7168,"tokenizer":"DeepSeek","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.75,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:55.561948+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3.1::base","model_id":"deepseek/deepseek-chat-v3.1","canonical_id":"deepseek/deepseek-chat-v3.1","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3.1","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":7168,"tokenizer":"DeepSeek","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.75,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:32:00.841614+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-4o-mini-2024-07-18::tools","model_id":"openai/gpt-4o-mini-2024-07-18","canonical_id":"openai/gpt-4o-mini-2024-07-18","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini (2024-07-18)","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:12.523267+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::cohere/command-r-08-2024::tools","model_id":"cohere/command-r-08-2024","canonical_id":"cohere/command-r-08-2024","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Cohere: Command R (08-2024)","org":"cohere","org_name":"Cohere","country":"Canada","city":"Toronto, ON","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":4000,"tokenizer":"Cohere","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Command series. Enterprise-focused. Strong RAG and retrieval performance. Canadian AI lab.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:09.477396+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::cohere/command-r-08-2024::structured","model_id":"cohere/command-r-08-2024","canonical_id":"cohere/command-r-08-2024","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Cohere: Command R (08-2024)","org":"cohere","org_name":"Cohere","country":"Canada","city":"Toronto, ON","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":4000,"tokenizer":"Cohere","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Command series. Enterprise-focused. Strong RAG and retrieval performance. Canadian AI lab.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":2906,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.592294+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::cohere/command-r-08-2024::low_temp","model_id":"cohere/command-r-08-2024","canonical_id":"cohere/command-r-08-2024","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Cohere: Command R (08-2024)","org":"cohere","org_name":"Cohere","country":"Canada","city":"Toronto, ON","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":4000,"tokenizer":"Cohere","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Command series. Enterprise-focused. Strong RAG and retrieval performance. Canadian AI lab.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:34.832694+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::cohere/command-r-08-2024::base","model_id":"cohere/command-r-08-2024","canonical_id":"cohere/command-r-08-2024","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Cohere: Command R (08-2024)","org":"cohere","org_name":"Cohere","country":"Canada","city":"Toronto, ON","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":4000,"tokenizer":"Cohere","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Command series. Enterprise-focused. Strong RAG and retrieval performance. Canadian AI lab.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:31:49.858964+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::deepseek-ai/DeepSeek-R1-Distill-Qwen-7B::base","model_id":"deepseek-ai/DeepSeek-R1-Distill-Qwen-7B","canonical_id":"deepseek-ai/DeepSeek-R1-Distill-Qwen-7B","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek-R1-Distill-Qwen-7B","org":"deepseek-ai","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":true,"context_length":131072,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":7,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":940,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:40.592728+00:00","status_source":"live_traffic","claimed_capabilities":null,"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::EssentialAI/rnj-1-instruct::tools","model_id":"EssentialAI/rnj-1-instruct","canonical_id":"EssentialAI/rnj-1-instruct","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"rnj-1-instruct","org":"EssentialAI","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":32768,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:05.551651+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::openai/gpt-4o-mini-search-preview-2025-03-11::structured","model_id":"openai/gpt-4o-mini-search-preview","canonical_id":"openai/gpt-4o-mini-search-preview-2025-03-11","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini Search Preview","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","response_format","structured_outputs","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1178,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.490962+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","response_format","structured_outputs","web_search_options"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::EssentialAI/rnj-1-instruct::low_temp","model_id":"EssentialAI/rnj-1-instruct","canonical_id":"EssentialAI/rnj-1-instruct","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"rnj-1-instruct","org":"EssentialAI","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":32768,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:05:22.182601+00:00","last_latency_ms":364,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:52.175696+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-4o-mini::structured","model_id":"openai/gpt-4o-mini","canonical_id":"openai/gpt-4o-mini","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_completion_tokens","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":728,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.605534+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_completion_tokens","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"huggingface::EssentialAI/rnj-1-instruct::base","model_id":"EssentialAI/rnj-1-instruct","canonical_id":"EssentialAI/rnj-1-instruct","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"rnj-1-instruct","org":"EssentialAI","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":32768,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.15,"pricing_output_per_1m":0.15,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:03:27.882473+00:00","last_latency_ms":730,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:12:08.064258+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-next-2025-02-03::low_temp","model_id":"qwen/qwen3-coder-next","canonical_id":"qwen/qwen3-coder-next-2025-02-03","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder Next","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.7999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":5504,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.881952+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::upstage/solar-pro-3::tools","model_id":"upstage/solar-pro-3","canonical_id":"upstage/solar-pro-3","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Upstage: Solar Pro 3","org":"upstage","org_name":"Upstage","country":"South Korea","city":"Seoul","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Solar series. Korean AI lab. Strong Korean and English. Document AI expertise.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:14.976343+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::openai/gpt-4o-mini::low_temp","model_id":"openai/gpt-4o-mini","canonical_id":"openai/gpt-4o-mini","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_completion_tokens","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:45:39.265214+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_completion_tokens","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::upstage/solar-pro-3::tools+reasoning","model_id":"upstage/solar-pro-3","canonical_id":"upstage/solar-pro-3","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Upstage: Solar Pro 3","org":"upstage","org_name":"Upstage","country":"South Korea","city":"Seoul","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Solar series. Korean AI lab. Strong Korean and English. Document AI expertise.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1186,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.087236+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::upstage/solar-pro-3::base","model_id":"upstage/solar-pro-3","canonical_id":"upstage/solar-pro-3","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Upstage: Solar Pro 3","org":"upstage","org_name":"Upstage","country":"South Korea","city":"Seoul","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Solar series. Korean AI lab. Strong Korean and English. Document AI expertise.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1036,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.100209+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::upstage/solar-pro-3::low_temp","model_id":"upstage/solar-pro-3","canonical_id":"upstage/solar-pro-3","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Upstage: Solar Pro 3","org":"upstage","org_name":"Upstage","country":"South Korea","city":"Seoul","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Solar series. Korean AI lab. Strong Korean and English. Document AI expertise.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1451,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.892143+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::upstage/solar-pro-3::reasoning","model_id":"upstage/solar-pro-3","canonical_id":"upstage/solar-pro-3","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Upstage: Solar Pro 3","org":"upstage","org_name":"Upstage","country":"South Korea","city":"Seoul","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Solar series. Korean AI lab. Strong Korean and English. Document AI expertise.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":764,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.483283+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::upstage/solar-pro-3::structured","model_id":"upstage/solar-pro-3","canonical_id":"upstage/solar-pro-3","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Upstage: Solar Pro 3","org":"upstage","org_name":"Upstage","country":"South Korea","city":"Seoul","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Solar series. Korean AI lab. Strong Korean and English. Document AI expertise.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":804,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.886990+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_tokens","reasoning","response_format","structured_outputs","temperature","tool_choice","tools"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::openai/gpt-4o-mini::tools","model_id":"openai/gpt-4o-mini","canonical_id":"openai/gpt-4o-mini","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-4o-mini","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":128000,"max_completion_tokens":16384,"tokenizer":"GPT","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_completion_tokens","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:16.831277+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_completion_tokens","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p","web_search_options"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-next-2025-02-03::base","model_id":"qwen/qwen3-coder-next","canonical_id":"qwen/qwen3-coder-next-2025-02-03","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder Next","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":262144,"tokenizer":"Qwen","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.7999999999999999,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":293,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.084635+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwq-32b::base","model_id":"qwen/qwq-32b","canonical_id":"qwen/qwq-32b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: QwQ 32B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Qwen","pricing_input_per_1m":0.15,"pricing_output_per_1m":0.58,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","reasoning","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:14:15.087799+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","reasoning","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-35b-a3b-20260224::base","model_id":"qwen/qwen3.5-35b-a3b","canonical_id":"qwen/qwen3.5-35b-a3b-20260224","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-35B-A3B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.1625,"pricing_output_per_1m":1.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":35,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1961,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.984600+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-35b-a3b-20260224::low_temp","model_id":"qwen/qwen3.5-35b-a3b","canonical_id":"qwen/qwen3.5-35b-a3b-20260224","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-35B-A3B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.1625,"pricing_output_per_1m":1.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":35,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":275,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.798509+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-35b-a3b-20260224::reasoning","model_id":"qwen/qwen3.5-35b-a3b","canonical_id":"qwen/qwen3.5-35b-a3b-20260224","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-35B-A3B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.1625,"pricing_output_per_1m":1.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":35,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":14532,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.397869+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-35b-a3b-20260224::structured","model_id":"qwen/qwen3.5-35b-a3b","canonical_id":"qwen/qwen3.5-35b-a3b-20260224","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-35B-A3B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.1625,"pricing_output_per_1m":1.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":35,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1631,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.793644+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-35b-a3b-20260224::tools+reasoning","model_id":"qwen/qwen3.5-35b-a3b","canonical_id":"qwen/qwen3.5-35b-a3b-20260224","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-35B-A3B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.1625,"pricing_output_per_1m":1.3,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":35,"active_param_count_b":3,"is_moe":true,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1956,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:04.972523+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::thedrummer/rocinante-12b::base","model_id":"thedrummer/rocinante-12b","canonical_id":"thedrummer/rocinante-12b","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"TheDrummer: Rocinante 12B","org":"thedrummer","org_name":"TheDrummer (community)","country":"Unknown","city":"Unknown","org_type":"community","open_weights":null,"context_length":32768,"max_completion_tokens":32768,"tokenizer":"Qwen","pricing_input_per_1m":0.16999999999999998,"pricing_output_per_1m":0.43,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Community fine-tunes focused on creative and uncensored outputs.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:14:08.403353+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::thedrummer/rocinante-12b::low_temp","model_id":"thedrummer/rocinante-12b","canonical_id":"thedrummer/rocinante-12b","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"TheDrummer: Rocinante 12B","org":"thedrummer","org_name":"TheDrummer (community)","country":"Unknown","city":"Unknown","org_type":"community","open_weights":null,"context_length":32768,"max_completion_tokens":32768,"tokenizer":"Qwen","pricing_input_per_1m":0.16999999999999998,"pricing_output_per_1m":0.43,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Community fine-tunes focused on creative and uncensored outputs.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:46:28.779685+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::thedrummer/rocinante-12b::tools","model_id":"thedrummer/rocinante-12b","canonical_id":"thedrummer/rocinante-12b","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"TheDrummer: Rocinante 12B","org":"thedrummer","org_name":"TheDrummer (community)","country":"Unknown","city":"Unknown","org_type":"community","open_weights":null,"context_length":32768,"max_completion_tokens":32768,"tokenizer":"Qwen","pricing_input_per_1m":0.16999999999999998,"pricing_output_per_1m":0.43,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Community fine-tunes focused on creative and uncensored outputs.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T01:23:15.177431+00:00","status_source":"verify","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::thedrummer/rocinante-12b::structured","model_id":"thedrummer/rocinante-12b","canonical_id":"thedrummer/rocinante-12b","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"TheDrummer: Rocinante 12B","org":"thedrummer","org_name":"TheDrummer (community)","country":"Unknown","city":"Unknown","org_type":"community","open_weights":null,"context_length":32768,"max_completion_tokens":32768,"tokenizer":"Qwen","pricing_input_per_1m":0.16999999999999998,"pricing_output_per_1m":0.43,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":12,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Community fine-tunes focused on creative and uncensored outputs.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":582,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.590474+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-27b-20260224::reasoning","model_id":"qwen/qwen3.5-27b","canonical_id":"qwen/qwen3.5-27b-20260224","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-27B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.195,"pricing_output_per_1m":1.56,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":9500,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.400282+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-27b-20260224::base","model_id":"qwen/qwen3.5-27b","canonical_id":"qwen/qwen3.5-27b-20260224","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-27B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.195,"pricing_output_per_1m":1.56,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":9978,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.986235+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3.5-27b-20260224::low_temp","model_id":"qwen/qwen3.5-27b","canonical_id":"qwen/qwen3.5-27b-20260224","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3.5-27B","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":262144,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.195,"pricing_output_per_1m":1.56,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","video"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":10760,"first_unavailable_at":"2026-04-12T01:23:06.384978+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.799756+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-flash::base","model_id":"qwen/qwen3-coder-flash","canonical_id":"qwen/qwen3-coder-flash","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder Flash","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.195,"pricing_output_per_1m":0.975,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","presence_penalty","response_format","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T14:14:18.914182+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","presence_penalty","response_format","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-flash::low_temp","model_id":"qwen/qwen3-coder-flash","canonical_id":"qwen/qwen3-coder-flash","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder Flash","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.195,"pricing_output_per_1m":0.975,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","presence_penalty","response_format","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:46:33.955017+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","presence_penalty","response_format","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::qwen/qwen3-coder-flash::tools","model_id":"qwen/qwen3-coder-flash","canonical_id":"qwen/qwen3-coder-flash","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen: Qwen3 Coder Flash","org":"qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":null,"context_length":1000000,"max_completion_tokens":65536,"tokenizer":"Qwen3","pricing_input_per_1m":0.195,"pricing_output_per_1m":0.975,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["max_tokens","presence_penalty","response_format","seed","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:27.562379+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","presence_penalty","response_format","seed","temperature","tool_choice","tools","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-4-fast::tools+reasoning","model_id":"x-ai/grok-4-fast","canonical_id":"x-ai/grok-4-fast","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":2811,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.298881+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-code-fast-1::base","model_id":"x-ai/grok-code-fast-1","canonical_id":"x-ai/grok-code-fast-1","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok Code Fast 1","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":256000,"max_completion_tokens":10000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:32:18.456627+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::x-ai/grok-code-fast-1::low_temp","model_id":"x-ai/grok-code-fast-1","canonical_id":"x-ai/grok-code-fast-1","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok Code Fast 1","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":256000,"max_completion_tokens":10000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:46:46.974307+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::x-ai/grok-code-fast-1::reasoning","model_id":"x-ai/grok-code-fast-1","canonical_id":"x-ai/grok-code-fast-1","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok Code Fast 1","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":256000,"max_completion_tokens":10000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1983,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.698675+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-code-fast-1::structured","model_id":"x-ai/grok-code-fast-1","canonical_id":"x-ai/grok-code-fast-1","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok Code Fast 1","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":256000,"max_completion_tokens":10000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":6268,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.719058+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-code-fast-1::tools","model_id":"x-ai/grok-code-fast-1","canonical_id":"x-ai/grok-code-fast-1","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok Code Fast 1","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":256000,"max_completion_tokens":10000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:44.998548+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3-0324::tools+reasoning","model_id":"deepseek/deepseek-chat-v3-0324","canonical_id":"deepseek/deepseek-chat-v3-0324","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3 0324","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":163840,"max_completion_tokens":null,"tokenizer":"DeepSeek","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.77,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":2016,"first_unavailable_at":null,"arch_id":"deepseek-v3.2","decoder_type":"Sparse MoE","attention":"MLA with DeepSeek Sparse Attention","arch_highlight":"An evolutionary update focused on efficiency rather than a new base layout.","tech_report_url":"https://arxiv.org/pdf/2512.02556","hf_config_url":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.610979+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3-0324::tools","model_id":"deepseek/deepseek-chat-v3-0324","canonical_id":"deepseek/deepseek-chat-v3-0324","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3 0324","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":163840,"max_completion_tokens":null,"tokenizer":"DeepSeek","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.77,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"deepseek-v3.2","decoder_type":"Sparse MoE","attention":"MLA with DeepSeek Sparse Attention","arch_highlight":"An evolutionary update focused on efficiency rather than a new base layout.","tech_report_url":"https://arxiv.org/pdf/2512.02556","hf_config_url":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:40.402529+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3-0324::structured","model_id":"deepseek/deepseek-chat-v3-0324","canonical_id":"deepseek/deepseek-chat-v3-0324","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3 0324","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":163840,"max_completion_tokens":null,"tokenizer":"DeepSeek","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.77,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1748,"first_unavailable_at":null,"arch_id":"deepseek-v3.2","decoder_type":"Sparse MoE","attention":"MLA with DeepSeek Sparse Attention","arch_highlight":"An evolutionary update focused on efficiency rather than a new base layout.","tech_report_url":"https://arxiv.org/pdf/2512.02556","hf_config_url":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.484892+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3-0324::reasoning","model_id":"deepseek/deepseek-chat-v3-0324","canonical_id":"deepseek/deepseek-chat-v3-0324","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3 0324","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":163840,"max_completion_tokens":null,"tokenizer":"DeepSeek","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.77,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":3353,"first_unavailable_at":null,"arch_id":"deepseek-v3.2","decoder_type":"Sparse MoE","attention":"MLA with DeepSeek Sparse Attention","arch_highlight":"An evolutionary update focused on efficiency rather than a new base layout.","tech_report_url":"https://arxiv.org/pdf/2512.02556","hf_config_url":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.894122+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3-0324::low_temp","model_id":"deepseek/deepseek-chat-v3-0324","canonical_id":"deepseek/deepseek-chat-v3-0324","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3 0324","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":163840,"max_completion_tokens":null,"tokenizer":"DeepSeek","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.77,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"deepseek-v3.2","decoder_type":"Sparse MoE","attention":"MLA with DeepSeek Sparse Attention","arch_highlight":"An evolutionary update focused on efficiency rather than a new base layout.","tech_report_url":"https://arxiv.org/pdf/2512.02556","hf_config_url":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:46:36.136121+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::deepseek/deepseek-chat-v3-0324::base","model_id":"deepseek/deepseek-chat-v3-0324","canonical_id":"deepseek/deepseek-chat-v3-0324","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"DeepSeek: DeepSeek V3 0324","org":"deepseek","org_name":"DeepSeek","country":"China","city":"Hangzhou","org_type":"startup","open_weights":null,"context_length":163840,"max_completion_tokens":null,"tokenizer":"DeepSeek","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.77,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logit_bias","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"DeepSeek-V/R series. Known for breakthrough cost efficiency. Strong code and math. Unique GRPO training (no RLHF). Open weights.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":"deepseek-v3.2","decoder_type":"Sparse MoE","attention":"MLA with DeepSeek Sparse Attention","arch_highlight":"An evolutionary update focused on efficiency rather than a new base layout.","tech_report_url":"https://arxiv.org/pdf/2512.02556","hf_config_url":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/config.json","variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:32:14.276800+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logit_bias","logprobs","max_tokens","min_p","presence_penalty","reasoning","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::x-ai/grok-4-fast::low_temp","model_id":"x-ai/grok-4-fast","canonical_id":"x-ai/grok-4-fast","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:46:37.987375+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/ministral-14b-2512::low_temp","model_id":"mistralai/ministral-14b-2512","canonical_id":"mistralai/ministral-14b-2512","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 14B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":14,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":302,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.984609+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::mistralai/ministral-14b-2512::base","model_id":"mistralai/ministral-14b-2512","canonical_id":"mistralai/ministral-14b-2512","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 14B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":14,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":428,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.222835+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::x-ai/grok-4-fast::reasoning","model_id":"x-ai/grok-4-fast","canonical_id":"x-ai/grok-4-fast","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1280,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.687574+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-4.1-fast::reasoning","model_id":"x-ai/grok-4.1-fast","canonical_id":"x-ai/grok-4.1-fast","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4.1 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":1364,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.590379+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-4.1-fast::structured","model_id":"x-ai/grok-4.1-fast","canonical_id":"x-ai/grok-4.1-fast","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4.1 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":4311,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.008896+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-4.1-fast::tools","model_id":"x-ai/grok-4.1-fast","canonical_id":"x-ai/grok-4.1-fast","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4.1 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:34.373628+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-4.1-fast::base","model_id":"x-ai/grok-4.1-fast","canonical_id":"x-ai/grok-4.1-fast","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4.1 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1930,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.296725+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::prime-intellect/intellect-3-20251126::tools+reasoning","model_id":"prime-intellect/intellect-3","canonical_id":"prime-intellect/intellect-3-20251126","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"Prime Intellect: INTELLECT-3","org":"prime-intellect","org_name":"Prime Intellect","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.1,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Intellect series. Decentralized/distributed training focus. Open research ethos.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":1345,"first_unavailable_at":"2026-04-12T07:27:33.073098+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.185601+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-4.1-fast::low_temp","model_id":"x-ai/grok-4.1-fast","canonical_id":"x-ai/grok-4.1-fast","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4.1 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:46:43.560860+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::x-ai/grok-4.1-fast::tools+reasoning","model_id":"x-ai/grok-4.1-fast","canonical_id":"x-ai/grok-4.1-fast","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4.1 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":2972,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.190293+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::mistralai/ministral-14b-2512::structured","model_id":"mistralai/ministral-14b-2512","canonical_id":"mistralai/ministral-14b-2512","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 14B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":14,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":338,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.916953+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-4-fast::structured","model_id":"x-ai/grok-4-fast","canonical_id":"x-ai/grok-4-fast","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":3263,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.117942+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-code-fast-1::tools+reasoning","model_id":"x-ai/grok-code-fast-1","canonical_id":"x-ai/grok-code-fast-1","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok Code Fast 1","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":256000,"max_completion_tokens":10000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":2447,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.312594+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::x-ai/grok-4-fast::base","model_id":"x-ai/grok-4-fast","canonical_id":"x-ai/grok-4-fast","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:32:20.693446+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-01::base","model_id":"minimax/minimax-01","canonical_id":"minimax/minimax-01","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax-01","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":1000192,"max_completion_tokens":1000192,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.1,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","temperature","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:28:44.380880+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:00:31.954166+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::minimax/minimax-01::low_temp","model_id":"minimax/minimax-01","canonical_id":"minimax/minimax-01","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"MiniMax: MiniMax-01","org":"minimax","org_name":"MiniMax","country":"China","city":"Shanghai","org_type":"startup","open_weights":null,"context_length":1000192,"max_completion_tokens":1000192,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.1,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["max_tokens","temperature","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"MiniMax series. Known for very long context windows. Strong Chinese language.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T07:29:41.375674+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T08:46:49.298403+00:00","status_source":"live_traffic","claimed_capabilities":["max_tokens","temperature","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-5.4-nano-20260317::reasoning","model_id":"openai/gpt-5.4-nano","canonical_id":"openai/gpt-5.4-nano-20260317","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-5.4 Nano","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":400000,"max_completion_tokens":128000,"tokenizer":"GPT","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.25,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["file","image","text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_completion_tokens","max_tokens","reasoning","response_format","seed","structured_outputs","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":624,"first_unavailable_at":"2026-04-12T07:26:41.196900+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.992568+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_completion_tokens","max_tokens","reasoning","response_format","seed","structured_outputs","tool_choice","tools"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::mistralai/mistral-saba-2502::structured","model_id":"mistralai/mistral-saba","canonical_id":"mistralai/mistral-saba-2502","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Saba","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":32768,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":473,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.496650+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","presence_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::mistralai/ministral-14b-2512::tools","model_id":"mistralai/ministral-14b-2512","canonical_id":"mistralai/ministral-14b-2512","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Mistral: Ministral 3 14B 2512","org":"mistralai","org_name":"Mistral AI","country":"France","city":"Paris","org_type":"startup","open_weights":null,"context_length":262144,"max_completion_tokens":null,"tokenizer":"Mistral","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.19999999999999998,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":14,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Mistral/Mixtral series. Strong European language coverage. Open weights for smaller models. Efficient MoE architecture.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:30.067132+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","logprobs","max_tokens","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"openrouter::prime-intellect/intellect-3-20251126::base","model_id":"prime-intellect/intellect-3","canonical_id":"prime-intellect/intellect-3-20251126","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Prime Intellect: INTELLECT-3","org":"prime-intellect","org_name":"Prime Intellect","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.1,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Intellect series. Decentralized/distributed training focus. Open research ethos.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1288,"first_unavailable_at":"2026-04-12T01:22:53.086255+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.292554+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::openai/gpt-5.4-nano-20260317::tools+reasoning","model_id":"openai/gpt-5.4-nano","canonical_id":"openai/gpt-5.4-nano-20260317","config_key":"tools+reasoning","variant_key":"tool_use_reasoning","interface":"openrouter","first_seen_at":"2026-04-01T05:02:00.852786+00:00","inference_config":{"tools":true,"reasoning":{"effort":"high"},"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use + Thinking)","variant_note":"Tool call with reasoning; maximum capability mode","use_tools":true,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":null,"is_claimed_valid":true,"is_retired":false,"name":"OpenAI: GPT-5.4 Nano","org":"openai","org_name":"OpenAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":400000,"max_completion_tokens":128000,"tokenizer":"GPT","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.25,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["file","image","text"],"output_modalities":["text"],"supported_parameters":["include_reasoning","max_completion_tokens","max_tokens","reasoning","response_format","seed","structured_outputs","tool_choice","tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"GPT series. Heavy English/Western web corpus. Strong coding and reasoning. RLHF alignment.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-04-01T05:02:00.852786+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:59:25.545687+00:00","last_latency_ms":646,"first_unavailable_at":"2026-04-12T07:26:41.595216+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:00:05.705879+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","max_completion_tokens","max_tokens","reasoning","response_format","seed","structured_outputs","tool_choice","tools"],"required_capabilities":null,"verified_capabilities":["reasoning","tools"],"route_count":0},{"variant_id":"openrouter::prime-intellect/intellect-3-20251126::reasoning","model_id":"prime-intellect/intellect-3","canonical_id":"prime-intellect/intellect-3-20251126","config_key":"reasoning","variant_key":"reasoning","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":{"effort":"medium"},"temperature":1,"tool_choice":null,"response_format":null,"max_completion_tokens":16384},"label_suffix":"(Thinking)","variant_note":"Extended thinking / chain-of-thought reasoning enabled","use_tools":false,"use_reasoning":true,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Prime Intellect: INTELLECT-3","org":"prime-intellect","org_name":"Prime Intellect","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.1,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Intellect series. Decentralized/distributed training focus. Open research ethos.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:58:42.547313+00:00","last_latency_ms":892,"first_unavailable_at":"2026-04-12T07:27:32.034342+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:59:24.583566+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["reasoning"],"verified_capabilities":["reasoning"],"route_count":0},{"variant_id":"openrouter::prime-intellect/intellect-3-20251126::structured","model_id":"prime-intellect/intellect-3","canonical_id":"prime-intellect/intellect-3-20251126","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"Prime Intellect: INTELLECT-3","org":"prime-intellect","org_name":"Prime Intellect","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":131072,"max_completion_tokens":131072,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":1.1,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Intellect series. Decentralized/distributed training focus. Open research ethos.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":1209,"first_unavailable_at":"2026-04-12T01:23:16.456393+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:57.002153+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","include_reasoning","max_tokens","presence_penalty","reasoning","repetition_penalty","response_format","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::allenai/olmo-3.1-32b-instruct-20251215::structured","model_id":"allenai/olmo-3.1-32b-instruct","canonical_id":"allenai/olmo-3.1-32b-instruct-20251215","config_key":"structured","variant_key":"structured_output","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"AllenAI: Olmo 3.1 32B Instruct","org":"allenai","org_name":"Allen Institute for AI","country":"USA","city":"Seattle, WA","org_type":"non-profit","open_weights":null,"context_length":65536,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"OLMo series. Fully open training data (Dolma). Strong research reproducibility focus. Non-profit academic lab.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":245,"first_unavailable_at":"2026-04-01T05:02:16.300317+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.892215+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":0},{"variant_id":"openrouter::allenai/olmo-3.1-32b-instruct-20251215::low_temp","model_id":"allenai/olmo-3.1-32b-instruct","canonical_id":"allenai/olmo-3.1-32b-instruct-20251215","config_key":"low_temp","variant_key":"completion_low_temp","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"AllenAI: Olmo 3.1 32B Instruct","org":"allenai","org_name":"Allen Institute for AI","country":"USA","city":"Seattle, WA","org_type":"non-profit","open_weights":null,"context_length":65536,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"OLMo series. Fully open training data (Dolma). Strong research reproducibility focus. Non-profit academic lab.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":462,"first_unavailable_at":"2026-04-01T05:02:16.300317+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.900245+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::allenai/olmo-3.1-32b-instruct-20251215::base","model_id":"allenai/olmo-3.1-32b-instruct","canonical_id":"allenai/olmo-3.1-32b-instruct-20251215","config_key":"base","variant_key":"completion","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"AllenAI: Olmo 3.1 32B Instruct","org":"allenai","org_name":"Allen Institute for AI","country":"USA","city":"Seattle, WA","org_type":"non-profit","open_weights":null,"context_length":65536,"max_completion_tokens":null,"tokenizer":"Other","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.6,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":32,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"OLMo series. Fully open training data (Dolma). Strong research reproducibility focus. Non-profit academic lab.","is_alias":true,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":426,"first_unavailable_at":"2026-04-01T05:02:16.300317+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:42.110879+00:00","status_source":"live_traffic","claimed_capabilities":["frequency_penalty","max_tokens","min_p","presence_penalty","repetition_penalty","response_format","seed","stop","structured_outputs","temperature","tool_choice","tools","top_k","top_p"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"openrouter::x-ai/grok-4-fast::tools","model_id":"x-ai/grok-4-fast","canonical_id":"x-ai/grok-4-fast","config_key":"tools","variant_key":"tool_use","interface":"openrouter","first_seen_at":"2026-03-13T04:16:17.881000+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_claimed_valid":true,"is_retired":false,"name":"xAI: Grok 4 Fast","org":"x-ai","org_name":"xAI","country":"USA","city":"San Francisco, CA","org_type":"startup","open_weights":null,"context_length":2000000,"max_completion_tokens":30000,"tokenizer":"Grok","pricing_input_per_1m":0.19999999999999998,"pricing_output_per_1m":0.5,"tags":"text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text","image","file"],"output_modalities":["text"],"supported_parameters":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Grok series. Trained on X (Twitter) data. Access to real-time information. Unfiltered/edgy persona.","is_alias":false,"source":"openrouter","model_first_seen_at":"2026-03-15T23:31:51.523000+00:00","model_last_seen_at":"2026-04-20T06:53:15.786763+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:23:08.011605+00:00","last_latency_ms":null,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:37.475438+00:00","status_source":"live_traffic","claimed_capabilities":["include_reasoning","logprobs","max_tokens","reasoning","response_format","seed","structured_outputs","temperature","tool_choice","tools","top_logprobs","top_p"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-Coder-Next::tools","model_id":"Qwen/Qwen3-Coder-Next","canonical_id":"Qwen/Qwen3-Coder-Next","config_key":"tools","variant_key":"tool_use","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":true,"reasoning":null,"temperature":1,"tool_choice":"required","response_format":null},"label_suffix":"(Tool Use)","variant_note":"Response via tool call; forces structured output through function calling","use_tools":true,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Coder-Next","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.2,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T01:10:18.669859+00:00","last_latency_ms":1667,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":null,"status_checked_at":"2026-04-12T09:39:47.049395+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["tools"],"verified_capabilities":["tools"],"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-Coder-Next::search","model_id":"Qwen/Qwen3-Coder-Next","canonical_id":"Qwen/Qwen3-Coder-Next","config_key":"search","variant_key":"search","interface":"huggingface","first_seen_at":"2026-04-13T19:43:59.804778+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null,"web_search_options":{"search_context_size":"medium"}},"label_suffix":"(Search)","variant_note":"Flags: search","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":true,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Coder-Next","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.2,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-13T20:36:35.845209+00:00","last_latency_ms":1180,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-13T20:37:23.289371+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":["web_search"],"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-Coder-Next::low_temp","model_id":"Qwen/Qwen3-Coder-Next","canonical_id":"Qwen/Qwen3-Coder-Next","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Coder-Next","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.2,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1477,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:32.697396+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::Qwen/Qwen3-Coder-Next::base","model_id":"Qwen/Qwen3-Coder-Next","canonical_id":"Qwen/Qwen3-Coder-Next","config_key":"base","variant_key":"completion","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":null},"label_suffix":"","variant_note":"Plain text completion, default temperature","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-Coder-Next","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":262144,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.2,"pricing_output_per_1m":1.5,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":null,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["tools"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:45:58.092369+00:00","last_latency_ms":1078,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:46:41.822998+00:00","status_source":"live_traffic","claimed_capabilities":["tools"],"required_capabilities":null,"verified_capabilities":null,"route_count":1},{"variant_id":"huggingface::aisingapore/Gemma-SEA-LION-v4-27B-IT::low_temp","model_id":"aisingapore/Gemma-SEA-LION-v4-27B-IT","canonical_id":"aisingapore/Gemma-SEA-LION-v4-27B-IT","config_key":"low_temp","variant_key":"completion_low_temp","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":0.2,"tool_choice":null,"response_format":null},"label_suffix":"(Low Temp)","variant_note":"Plain text completion, temperature=0.2 for deterministic comparison","use_tools":false,"use_reasoning":false,"use_structured_output":false,"use_web_search":false,"use_low_temp":true,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Gemma-SEA-LION-v4-27B-IT","org":"aisingapore","org_name":null,"country":null,"city":null,"org_type":null,"open_weights":true,"context_length":null,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.2,"pricing_output_per_1m":0.4,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":27,"active_param_count_b":null,"is_moe":false,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":["structured_outputs"],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":null,"is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T13:46:43.084470+00:00","last_latency_ms":1523,"first_unavailable_at":null,"arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T13:47:31.902023+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs"],"required_capabilities":null,"verified_capabilities":null,"route_count":0},{"variant_id":"huggingface::Qwen/Qwen3-235B-A22B::structured","model_id":"Qwen/Qwen3-235B-A22B","canonical_id":"Qwen/Qwen3-235B-A22B","config_key":"structured","variant_key":"structured_output","interface":"huggingface","first_seen_at":"2026-04-01T05:48:12.876428+00:00","inference_config":{"tools":null,"reasoning":null,"temperature":1,"tool_choice":null,"response_format":"json_schema"},"label_suffix":"(Structured Output)","variant_note":"Response via response_format=json_schema; schema-validated JSON without tools","use_tools":false,"use_reasoning":false,"use_structured_output":true,"use_web_search":false,"use_low_temp":false,"variant_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_claimed_valid":true,"is_retired":false,"name":"Qwen3-235B-A22B","org":"Qwen","org_name":"Alibaba DAMO Academy","country":"China","city":"Hangzhou","org_type":"big-tech","open_weights":true,"context_length":40960,"max_completion_tokens":null,"tokenizer":null,"pricing_input_per_1m":0.2,"pricing_output_per_1m":0.8,"tags":"open_weight text_generation","release_date":null,"expiration_date":null,"param_count_b":235,"active_param_count_b":22,"is_moe":true,"specialization":null,"input_modalities":["text"],"output_modalities":["text"],"supported_parameters":[],"rate_limit_rpm":null,"rate_limit_rpd":null,"rate_limit_tpm":null,"rate_limit_source":null,"provenance_notes":"Qwen series. Extremely strong Chinese coverage. Competitive English. Strong coding and math. Released many open-weight variants.","is_alias":false,"source":"huggingface","model_first_seen_at":"2026-04-01T05:48:12.876428+00:00","model_last_seen_at":"2026-04-20T06:54:09.379813+00:00","is_available":true,"unavailable_reason":null,"last_checked_at":"2026-04-12T14:01:59.872642+00:00","last_latency_ms":2350,"first_unavailable_at":"2026-04-12T13:14:11.279531+00:00","arch_id":null,"decoder_type":null,"attention":null,"arch_highlight":null,"tech_report_url":null,"hf_config_url":null,"variant_status":"available","reason_code":null,"reason_detail":null,"status_http_status":200,"status_checked_at":"2026-04-12T14:07:56.613042+00:00","status_source":"live_traffic","claimed_capabilities":["structured_outputs","tools"],"required_capabilities":["structured_outputs"],"verified_capabilities":["structured_outputs"],"route_count":2}]}