[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"api:\u002Fnav:{}":3,"api:\u002Fbenchmarks\u002Fautomationbench:{\"effort\":\"top\",\"limit\":100,\"page\":1}":104},{"login_code_required":4,"data_date":5,"vendors":6,"categories":39,"topics":55,"model_count":72,"benchmark_count":73,"models":74,"benchmarks":79},false,"2026-10-27",[7,15,23,31],{"name":8,"slug":9,"models":10},"阿里巴巴","alibaba",[11],{"name":12,"slug":13,"short_name":12,"overall_score":14},"Qwen3.8-27B","qwen3-8-27b","47.576875000000000000000000",{"name":16,"slug":17,"models":18},"深度求索","deepseek",[19],{"name":20,"slug":21,"short_name":20,"overall_score":22},"DeepSeek-V4-Pro","deepseek-v4-pro","65.030000000000000000000000",{"name":24,"slug":25,"models":26},"月之暗面","moonshot",[27],{"name":28,"slug":29,"short_name":28,"overall_score":30},"Kimi-K3","kimi-k3","57.474000000000000000000000",{"name":32,"slug":33,"models":34},"智谱","zhipu",[35],{"name":36,"slug":37,"short_name":36,"overall_score":38},"GLM-5.3","glm-5-3","61.067500000000000000000000",[40,44,48,51],{"name":41,"slug":42,"benchmark_count":43},"体验感","experience",2,{"name":45,"slug":46,"benchmark_count":47},"认知能力","cognition",5,{"name":49,"slug":50,"benchmark_count":43},"Agentic 能力","agentic",{"name":52,"slug":53,"benchmark_count":54},"应用领域","domain",1,[56,60,64,68],{"slug":57,"name":58,"question":59},"quant","量化 vs 原精度","量化之后能力掉多少",{"slug":61,"name":62,"question":63},"selfhost","自部署 vs 官方 API","同一份权重，自己部署和调官方 API，分数一致吗",{"slug":65,"name":66,"question":67},"size","同家族尺寸","同一代模型，小尺寸比旗舰差多少",{"slug":69,"name":70,"question":71},"harness","多 Harness","同一个模型换一个 harness，agentic 分数变多少",4,8,[75,76,77,78],{"name":20,"slug":21},{"name":36,"slug":37},{"name":28,"slug":29},{"name":12,"slug":13},[80,83,86,89,92,95,98,101],{"name":81,"slug":82},"AA-LCR","aa-lcr",{"name":84,"slug":85},"AA-Omniscience","aa-omniscience",{"name":87,"slug":88},"AutomationBench","automationbench",{"name":90,"slug":91},"CL-bench","cl-bench",{"name":93,"slug":94},"HLE","hle",{"name":96,"slug":97},"MMLU-Pro","mmlu-pro",{"name":99,"slug":100},"SimpleQA Verified","simpleqa-verified",{"name":102,"slug":103},"SWE-bench Pro","swe-bench-pro",{"benchmark":105,"updated_at":116,"rows":117,"total":72,"categories":223,"tabs":226,"settings":228},{"id":43,"name":87,"slug":88,"license":106,"version":107,"is_active":108,"num_cases":109,"parent_id":107,"score_max":110,"score_min":111,"created_at":112,"has_bundle":108,"is_agentic":108,"sort_order":113,"updated_at":112,"description":107,"metric_name":114,"homepage_url":107,"license_note":107,"scoring_method":107,"composite_weight":110,"higher_is_better":108,"bundle_size_bytes":115},"public",null,true,600,"1.0000","0.0000","2026-10-28T09:47:38.906402+08:00",0,"Pass Rate",2028958067,"2026-10-26",[118,151,176,200],{"id":119,"rank":54,"notes":107,"score":120,"access":121,"effort":122,"family":36,"is_tier":108,"metrics":123,"cost_usd":107,"hardware":107,"model_id":54,"provider":107,"settings":124,"data_date":131,"num_cases":109,"precision":107,"score_std":107,"categories":132,"harness_id":107,"is_primary":108,"model_name":36,"model_slug":37,"output_tps":107,"short_name":36,"variant_id":73,"effort_rank":142,"input_price":107,"num_samples":54,"vendor_name":32,"vendor_slug":33,"architecture":107,"benchmark_id":43,"evaluated_at":143,"harness_name":107,"harness_slug":107,"input_tokens":144,"output_price":107,"overall_rank":43,"ranked_count":72,"release_date":145,"ttft_seconds":107,"variant_name":146,"variant_slug":147,"vendor_order":113,"weights_repo":107,"missing_count":43,"output_tokens":148,"overall_score":38,"parent_run_id":107,"context_window":107,"is_primary_run":108,"total_cost_usd":107,"total_params_b":107,"weights_source":107,"active_params_b":107,"cache_hit_price":107,"cost_incomplete":108,"harness_version":107,"is_open_weights":108,"avg_case_seconds":149,"inference_engine":107,"reasoning_tokens":107,"cost_per_case_usd":107,"trajectory_bundle_size_bytes":150},41,"0.5250","api","max",{},{"top_p":125,"recipe":107,"harness":107,"judge_model":107,"num_samples":54,"temperature":126,"extra_settings":127,"scoring_method":107,"harness_version":107,"framework_version":129,"max_output_tokens":130},"1.000","0.00",{"top_k":128},-1,"OPT 1478395",32768,"2026-10-23",[133,135,138,140],{"name":41,"rank":54,"slug":42,"score":134,"expected":43,"available":54,"ranked_count":72},"72.997032640949555000000000",{"name":45,"rank":136,"slug":46,"score":137,"expected":47,"available":72,"ranked_count":72},3,"59.363333333333333750000000",{"name":49,"rank":54,"slug":50,"score":139,"expected":43,"available":43,"ranked_count":72},"54.225000000000000000000000",{"name":52,"rank":54,"slug":53,"score":141,"expected":54,"available":54,"ranked_count":72},"52.500000000000000000000000",6,"2026-10-19",281739014,"2026-08-18","GLM-5.3 官方 API max","glm-5-3-api-max",6962692,"868.68",22105164,{"id":152,"rank":43,"notes":107,"score":153,"access":154,"effort":122,"family":155,"is_tier":108,"metrics":156,"cost_usd":107,"hardware":107,"model_id":72,"provider":107,"settings":157,"data_date":5,"num_cases":109,"precision":107,"score_std":107,"categories":161,"harness_id":107,"is_primary":108,"model_name":20,"model_slug":21,"output_tps":107,"short_name":20,"variant_id":136,"effort_rank":142,"input_price":107,"num_samples":54,"vendor_name":16,"vendor_slug":17,"architecture":107,"benchmark_id":43,"evaluated_at":116,"harness_name":107,"harness_slug":107,"input_tokens":169,"output_price":107,"overall_rank":54,"ranked_count":72,"release_date":170,"ttft_seconds":107,"variant_name":171,"variant_slug":172,"vendor_order":113,"weights_repo":107,"missing_count":136,"output_tokens":173,"overall_score":22,"parent_run_id":107,"context_window":107,"is_primary_run":108,"total_cost_usd":107,"total_params_b":107,"weights_source":107,"active_params_b":107,"cache_hit_price":107,"cost_incomplete":108,"harness_version":107,"is_open_weights":108,"avg_case_seconds":174,"inference_engine":107,"reasoning_tokens":107,"cost_per_case_usd":107,"trajectory_bundle_size_bytes":175},14,"0.4567","self_hosted","DeepSeek-V4",{},{"top_p":158,"recipe":107,"harness":107,"judge_model":107,"num_samples":54,"temperature":159,"extra_settings":160,"scoring_method":107,"harness_version":107,"framework_version":129,"max_output_tokens":130},"0.950","1.00",{"top_k":128},[162,164,166,168],{"name":41,"rank":72,"slug":42,"score":163,"expected":43,"available":54,"ranked_count":72},"2.857142857142860000000000",{"name":45,"rank":54,"slug":46,"score":165,"expected":47,"available":72,"ranked_count":72},"69.724166666666667500000000",{"name":49,"rank":136,"slug":50,"score":167,"expected":43,"available":54,"ranked_count":72},"45.670000000000000000000000",{"name":52,"rank":43,"slug":53,"score":167,"expected":54,"available":54,"ranked_count":72},704064732,"2026-08-13","DeepSeek-V4-Pro 自部署 max","deepseek-v4-pro-self-hosted-max",12837730,"970.09",32227494,{"id":177,"rank":136,"notes":107,"score":178,"access":154,"effort":179,"family":180,"is_tier":108,"metrics":181,"cost_usd":107,"hardware":107,"model_id":43,"provider":107,"settings":182,"data_date":185,"num_cases":109,"precision":107,"score_std":107,"categories":186,"harness_id":107,"is_primary":108,"model_name":12,"model_slug":13,"output_tps":107,"short_name":12,"variant_id":195,"effort_rank":47,"input_price":107,"num_samples":54,"vendor_name":8,"vendor_slug":9,"architecture":107,"benchmark_id":43,"evaluated_at":143,"harness_name":107,"harness_slug":107,"input_tokens":107,"output_price":107,"overall_rank":72,"ranked_count":72,"release_date":107,"ttft_seconds":107,"variant_name":196,"variant_slug":197,"vendor_order":113,"weights_repo":107,"missing_count":113,"output_tokens":198,"overall_score":14,"parent_run_id":107,"context_window":107,"is_primary_run":108,"total_cost_usd":107,"total_params_b":107,"weights_source":107,"active_params_b":107,"cache_hit_price":107,"cost_incomplete":108,"harness_version":107,"is_open_weights":108,"avg_case_seconds":107,"inference_engine":107,"reasoning_tokens":107,"cost_per_case_usd":107,"trajectory_bundle_size_bytes":199},114,"0.4150","xhigh","Qwen3.8",{},{"top_p":158,"recipe":107,"harness":107,"judge_model":107,"num_samples":54,"temperature":159,"extra_settings":183,"scoring_method":107,"harness_version":107,"framework_version":129,"max_output_tokens":130},{"top_k":184},20,"2026-10-22",[187,189,191,193],{"name":41,"rank":136,"slug":42,"score":188,"expected":43,"available":43,"ranked_count":72},"48.453230452674897000000000",{"name":45,"rank":72,"slug":46,"score":190,"expected":47,"available":47,"ranked_count":72},"45.520000000000000000000000",{"name":49,"rank":43,"slug":50,"score":192,"expected":43,"available":43,"ranked_count":72},"49.480000000000000000000000",{"name":52,"rank":136,"slug":53,"score":194,"expected":54,"available":54,"ranked_count":72},"41.500000000000000000000000",19,"Qwen3.8-27B 自部署 xhigh","qwen3-8-27b-self-hosted-xhigh",8589061,14739968,{"id":201,"rank":72,"notes":107,"score":202,"access":121,"effort":122,"family":28,"is_tier":108,"metrics":203,"cost_usd":107,"hardware":107,"model_id":136,"provider":107,"settings":204,"data_date":206,"num_cases":109,"precision":107,"score_std":107,"categories":207,"harness_id":107,"is_primary":108,"model_name":28,"model_slug":29,"output_tps":107,"short_name":28,"variant_id":215,"effort_rank":142,"input_price":107,"num_samples":54,"vendor_name":24,"vendor_slug":25,"architecture":107,"benchmark_id":43,"evaluated_at":206,"harness_name":107,"harness_slug":107,"input_tokens":216,"output_price":107,"overall_rank":136,"ranked_count":72,"release_date":217,"ttft_seconds":107,"variant_name":218,"variant_slug":219,"vendor_order":113,"weights_repo":107,"missing_count":136,"output_tokens":220,"overall_score":30,"parent_run_id":107,"context_window":107,"is_primary_run":108,"total_cost_usd":107,"total_params_b":107,"weights_source":107,"active_params_b":107,"cache_hit_price":107,"cost_incomplete":108,"harness_version":107,"is_open_weights":108,"avg_case_seconds":221,"inference_engine":107,"reasoning_tokens":107,"cost_per_case_usd":107,"trajectory_bundle_size_bytes":222},59,"0.0350",{},{"top_p":158,"recipe":107,"harness":107,"judge_model":107,"num_samples":54,"temperature":159,"extra_settings":205,"scoring_method":107,"harness_version":107,"framework_version":129,"max_output_tokens":130},{"top_k":128},"2026-10-24",[208,210,212,214],{"name":41,"rank":43,"slug":42,"score":209,"expected":43,"available":54,"ranked_count":72},"65.051903114186850000000000",{"name":45,"rank":43,"slug":46,"score":211,"expected":47,"available":72,"ranked_count":72},"67.050833333333332500000000",{"name":49,"rank":72,"slug":50,"score":213,"expected":43,"available":54,"ranked_count":72},"3.500000000000000000000000",{"name":52,"rank":72,"slug":53,"score":213,"expected":54,"available":54,"ranked_count":72},11,1389600376,"2026-07-16","Kimi-K3 官方 API max","kimi-k3-api-max",15364108,"4843.73",26581258,[224,225],{"name":49,"slug":50},{"name":52,"slug":53},[227],{"name":87,"slug":88,"num_cases":109},{"extra_settings":229,"framework_version":231,"harness":232,"harness_version":233,"judge_model":234,"max_output_tokens":235,"num_samples":236,"recipe":237,"scoring_method":238,"temperature":239,"top_p":240},{"value":107,"varies":108,"label":230},"各记录不同",{"value":129,"varies":4,"label":107},{"value":107,"varies":4,"label":107},{"value":107,"varies":4,"label":107},{"value":107,"varies":4,"label":107},{"value":130,"varies":4,"label":107},{"value":54,"varies":4,"label":107},{"value":107,"varies":4,"label":107},{"value":107,"varies":4,"label":107},{"value":107,"varies":108,"label":230},{"value":107,"varies":108,"label":230}]