{"source":"AIMap","schema":"benchmark/v1","updated":"2026-10-22","benchmark":{"id":"hle","name":"Humanity's Last Exam","category":"knowledge","unit":"%","higherIsBetter":true,"status":"active","sources":["epoch"],"primaryVariant":null,"url":"https://lastexam.ai/","year":2025,"desc":{"en":"About 2,500 expert-written questions across 100+ subjects from the Center for AI Safety and Scale AI, filtered to stump frontier models at creation time.","zh-hans":"由 Center for AI Safety 和 Scale AI 组织、约 2500 道专家出题，覆盖 100 多个学科，出题时专门筛掉了前沿模型能答对的题。"},"measures":{"en":"Accuracy (mostly exact-match answers). Data compiled by Epoch AI.","zh-hans":"准确率（以精确匹配为主）。数据由 Epoch AI 汇编。"},"limits":{"en":"Independent audits reported incorrect reference answers in some subjects; settings (tools, text-only) differ between reports.","zh-hans":"独立审查发现部分学科的标准答案有误；不同报告的设定（是否用工具、是否只做文本题）不同。"},"contamination":"low","contaminationNote":{"en":"Private held-out set exists; public questions carry a canary string.","zh-hans":"保留了私有题集；公开题带有防训练标记。"},"stats":{"results":51,"models":46,"latest":"2026-10-07","bySourceType":{"compiled":51}}},"sources":[{"id":"epoch","name":"Epoch AI Benchmarking Hub","license":"CC BY 4.0","policy":"open","attribution":"Epoch AI, 'Capabilities & benchmarking', https://epoch.ai/benchmarks (CC BY 4.0)","site":"https://epoch.ai/benchmarks"}],"results":[{"model":"gpt-6-astra","modelName":"GPT-6 Astra","provider":"OpenAI","rawName":"GPT-6 Astra","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":54.8,"unit":"%","rank":1,"rankComputedBySite":true,"date":"2026-09-03","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-fable-5-1","modelName":"Claude Fable 5.1","provider":"Anthropic","rawName":"Claude Fable 5.1","variant":"compiled","setting":"Compiled by Epoch AI; effort: xhigh","score":46.5,"unit":"%","rank":2,"rankComputedBySite":true,"date":"2026-09-01","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-3-1-pro","modelName":"Gemini 3.1 Pro","provider":"Google","rawName":"Gemini 3.1 Pro","variant":"compiled","setting":"Compiled by Epoch AI","score":46.44,"unit":"%","rank":3,"rankComputedBySite":true,"date":"2026-02-19","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-sonnet-5-5","modelName":"Claude Sonnet 5.5","provider":"Anthropic","rawName":"Claude Sonnet 5.5","variant":"compiled","setting":"Compiled by Epoch AI; effort: xhigh","score":46.12,"unit":"%","rank":4,"rankComputedBySite":true,"date":"2026-09-28","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-22"},{"model":"gemini-3-8-flash","modelName":"Gemini 3.8 Flash","provider":"Google","rawName":"Gemini 3.8 Flash","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown","score":44.52,"unit":"%","rank":5,"rankComputedBySite":true,"date":"2026-09-02","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-5-4-pro","modelName":"GPT-5.4 Pro","provider":"OpenAI","rawName":"GPT-5.4 Pro","variant":"compiled","setting":"Compiled by Epoch AI","score":44.32,"unit":"%","rank":6,"rankComputedBySite":true,"date":"2026-03-05","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"muse-spark","modelName":"Muse Spark","provider":"Meta","rawName":"Muse Spark","variant":"compiled","setting":"Compiled by Epoch AI","score":40.56,"unit":"%","rank":7,"rankComputedBySite":true,"date":"2026-04-08","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-3-pro","modelName":"Gemini 3 Pro","provider":"Google","rawName":"Gemini 3 Pro","variant":"compiled","setting":"Compiled by Epoch AI","score":37.52,"unit":"%","rank":8,"rankComputedBySite":true,"date":"2025-11-18","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-5-4","modelName":"GPT-5.4","provider":"OpenAI","rawName":"GPT-5.4","variant":"compiled","setting":"Compiled by Epoch AI; effort: xhigh","score":36.24,"unit":"%","rank":9,"rankComputedBySite":true,"date":"2026-03-05","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-opus-4-7","modelName":"Claude Opus 4.7","provider":"Anthropic","rawName":"Claude Opus 4.7","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown","score":36.2,"unit":"%","rank":10,"rankComputedBySite":true,"date":"2026-04-16","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-opus-4-6","modelName":"Claude Opus 4.6","provider":"Anthropic","rawName":"Claude Opus 4.6","variant":"compiled","setting":"Compiled by Epoch AI; effort: max","score":34.44,"unit":"%","rank":11,"rankComputedBySite":true,"date":"2026-02-05","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-5-pro","modelName":"GPT-5 Pro","provider":"OpenAI","rawName":"GPT-5 Pro","variant":"compiled","setting":"Compiled by Epoch AI","score":31.64,"unit":"%","rank":12,"rankComputedBySite":true,"date":"2025-10-07","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-haiku-5-5","modelName":"Claude Haiku 5.5","provider":"Anthropic","rawName":"Claude Haiku 5.5","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown","score":30.12,"unit":"%","rank":13,"rankComputedBySite":true,"date":"2026-10-07","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-22"},{"model":"gpt-5-2","modelName":"GPT-5.2","provider":"OpenAI","rawName":"GPT-5.2","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":27.8,"unit":"%","rank":14,"rankComputedBySite":true,"date":"2025-12-11","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-5","modelName":"GPT-5","provider":"OpenAI","rawName":"GPT-5","variant":"compiled","setting":"Compiled by Epoch AI; effort: high","score":25.32,"unit":"%","rank":15,"rankComputedBySite":true,"date":"2025-08-07","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-5","modelName":"GPT-5","provider":"OpenAI","rawName":"GPT-5","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":25.32,"unit":"%","rank":15,"rankComputedBySite":true,"date":"2025-08-07","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-opus-4-5","modelName":"Claude Opus 4.5","provider":"Anthropic","rawName":"Claude Opus 4.5","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":25.2,"unit":"%","rank":17,"rankComputedBySite":true,"date":"2025-11-24","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"kimi-k2-5","modelName":"Kimi K2.5","provider":"Moonshot AI","rawName":"Kimi K2.5","variant":"compiled","setting":"Compiled by Epoch AI","score":24.37,"unit":"%","rank":18,"rankComputedBySite":true,"date":"2026-01-27","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-5-1","modelName":"GPT-5.1","provider":"OpenAI","rawName":"GPT-5.1","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":23.68,"unit":"%","rank":19,"rankComputedBySite":true,"date":"2025-11-13","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-2-5-pro-jun-2025","modelName":"Gemini 2.5 Pro (Jun 2025)","provider":"Google","rawName":"Gemini 2.5 Pro (Jun 2025)","variant":"compiled","setting":"Compiled by Epoch AI; effort: Jun 2025","score":21.64,"unit":"%","rank":20,"rankComputedBySite":true,"date":"2025-06-05","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"o3","modelName":"o3","provider":"OpenAI","rawName":"o3","variant":"compiled","setting":"Compiled by Epoch AI; effort: high","score":20.32,"unit":"%","rank":21,"rankComputedBySite":true,"date":"2025-04-16","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-5-mini","modelName":"GPT-5 mini","provider":"OpenAI","rawName":"GPT-5 mini","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":19.44,"unit":"%","rank":22,"rankComputedBySite":true,"date":"2025-08-07","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"o3","modelName":"o3","provider":"OpenAI","rawName":"o3","variant":"compiled","setting":"Compiled by Epoch AI; effort: medium","score":19.2,"unit":"%","rank":23,"rankComputedBySite":true,"date":"2025-04-16","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-opus-4-6","modelName":"Claude Opus 4.6","provider":"Anthropic","rawName":"Claude Opus 4.6","variant":"compiled","setting":"Compiled by Epoch AI; effort: no thinking","score":19,"unit":"%","rank":24,"rankComputedBySite":true,"date":"2026-02-05","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-2-5-pro-mar-2025","modelName":"Gemini 2.5 Pro (Mar 2025)","provider":"Google","rawName":"Gemini 2.5 Pro (Mar 2025)","variant":"compiled","setting":"Compiled by Epoch AI; effort: Mar 2025","score":18.16,"unit":"%","rank":25,"rankComputedBySite":true,"date":"2025-03-25","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"o4-mini","modelName":"o4-mini","provider":"OpenAI","rawName":"o4-mini","variant":"compiled","setting":"Compiled by Epoch AI; effort: high","score":18.08,"unit":"%","rank":26,"rankComputedBySite":true,"date":"2025-04-16","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-2-5-pro-may-2025","modelName":"Gemini 2.5 Pro (May 2025)","provider":"Google","rawName":"Gemini 2.5 Pro (May 2025)","variant":"compiled","setting":"Compiled by Epoch AI; effort: Jun 2025","score":17.8,"unit":"%","rank":27,"rankComputedBySite":true,"date":"2025-05-06","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"o4-mini","modelName":"o4-mini","provider":"OpenAI","rawName":"o4-mini","variant":"compiled","setting":"Compiled by Epoch AI; effort: medium","score":14.28,"unit":"%","rank":28,"rankComputedBySite":true,"date":"2025-04-16","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-sonnet-4-5","modelName":"Claude Sonnet 4.5","provider":"Anthropic","rawName":"Claude Sonnet 4.5","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":13.72,"unit":"%","rank":29,"rankComputedBySite":true,"date":"2025-09-29","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-2-5-flash-apr-2025","modelName":"Gemini 2.5 Flash (Apr 2025)","provider":"Google","rawName":"Gemini 2.5 Flash (Apr 2025)","variant":"compiled","setting":"Compiled by Epoch AI; effort: Apr 2025","score":12.08,"unit":"%","rank":30,"rankComputedBySite":true,"date":"2025-04-17","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-opus-4-1","modelName":"Claude Opus 4.1","provider":"Anthropic","rawName":"Claude Opus 4.1","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":11.52,"unit":"%","rank":31,"rankComputedBySite":true,"date":"2025-08-05","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-2-5-flash-may-2025","modelName":"Gemini 2.5 Flash (May 2025)","provider":"Google","rawName":"Gemini 2.5 Flash (May 2025)","variant":"compiled","setting":"Compiled by Epoch AI","score":10.96,"unit":"%","rank":32,"rankComputedBySite":true,"date":"2025-05-20","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-opus-4","modelName":"Claude Opus 4","provider":"Anthropic","rawName":"Claude Opus 4","variant":"compiled","setting":"Compiled by Epoch AI","score":10.72,"unit":"%","rank":33,"rankComputedBySite":true,"date":"2025-05-22","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-3-1-flash-lite","modelName":"Gemini 3.1 Flash-Lite","provider":"Google","rawName":"Gemini 3.1 Flash-Lite","variant":"compiled","setting":"Compiled by Epoch AI","score":8.64,"unit":"%","rank":34,"rankComputedBySite":true,"date":"2026-03-03","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"glm-4-5","modelName":"GLM-4.5","provider":"Zhipu AI (Z.ai)","rawName":"GLM-4.5","variant":"compiled","setting":"Compiled by Epoch AI","score":8.32,"unit":"%","rank":35,"rankComputedBySite":true,"date":"2025-08-03","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"glm-4-5-air","modelName":"GLM-4.5-Air","provider":"Zhipu AI (Z.ai)","rawName":"GLM-4.5-Air","variant":"compiled","setting":"Compiled by Epoch AI","score":8.12,"unit":"%","rank":36,"rankComputedBySite":true,"date":"2025-07-20","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"o1-pro","modelName":"o1-pro","provider":"OpenAI","rawName":"o1-pro","variant":"compiled","setting":"Compiled by Epoch AI","score":8.12,"unit":"%","rank":36,"rankComputedBySite":true,"date":"2025-03-19","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-3-7-sonnet","modelName":"Claude 3.7 Sonnet","provider":"Anthropic","rawName":"Claude 3.7 Sonnet","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":8.04,"unit":"%","rank":38,"rankComputedBySite":true,"date":"2025-02-24","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"o1","modelName":"o1","provider":"OpenAI","rawName":"o1","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":7.96,"unit":"%","rank":39,"rankComputedBySite":true,"date":"2024-12-17","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-sonnet-4","modelName":"Claude Sonnet 4","provider":"Anthropic","rawName":"Claude Sonnet 4","variant":"compiled","setting":"Compiled by Epoch AI; effort: unknown thinking","score":7.76,"unit":"%","rank":40,"rankComputedBySite":true,"date":"2025-05-22","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-5-1","modelName":"GPT-5.1","provider":"OpenAI","rawName":"GPT-5.1","variant":"compiled","setting":"Compiled by Epoch AI; effort: no thinking","score":6.8,"unit":"%","rank":41,"rankComputedBySite":true,"date":"2025-11-13","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-2-0-flash-thinking-jan-2025","modelName":"Gemini 2.0 Flash Thinking (Jan 2025)","provider":"Google","rawName":"Gemini 2.0 Flash Thinking (Jan 2025)","variant":"compiled","setting":"Compiled by Epoch AI","score":6.56,"unit":"%","rank":42,"rankComputedBySite":true,"date":"2025-01-21","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"llama-4-maverick","modelName":"Llama 4 Maverick","provider":"Meta","rawName":"Llama 4 Maverick","variant":"compiled","setting":"Compiled by Epoch AI","score":5.68,"unit":"%","rank":43,"rankComputedBySite":true,"date":"2025-04-05","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-4-5","modelName":"GPT-4.5","provider":"OpenAI","rawName":"GPT-4.5","variant":"compiled","setting":"Compiled by Epoch AI; effort: Feb 2025","score":5.44,"unit":"%","rank":44,"rankComputedBySite":true,"date":"2025-02-27","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-4-1","modelName":"GPT-4.1","provider":"OpenAI","rawName":"GPT-4.1","variant":"compiled","setting":"Compiled by Epoch AI","score":5.4,"unit":"%","rank":45,"rankComputedBySite":true,"date":"2025-04-14","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gemini-1-5-pro-sept-2024","modelName":"Gemini 1.5 Pro (Sept 2024)","provider":"Google","rawName":"Gemini 1.5 Pro (Sept 2024)","variant":"compiled","setting":"Compiled by Epoch AI","score":4.6,"unit":"%","rank":46,"rankComputedBySite":true,"date":"2024-09-24","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"mistral-medium-3","modelName":"Mistral Medium 3","provider":"Mistral AI","rawName":"Mistral Medium 3","variant":"compiled","setting":"Compiled by Epoch AI","score":4.52,"unit":"%","rank":47,"rankComputedBySite":true,"date":"2025-05-07","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"amazon-nova-pro","modelName":"Amazon Nova Pro","provider":"Amazon","rawName":"Amazon Nova Pro","variant":"compiled","setting":"Compiled by Epoch AI","score":4.4,"unit":"%","rank":48,"rankComputedBySite":true,"date":"2024-12-03","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"claude-3-5-sonnet-october-2024","modelName":"Claude 3.5 Sonnet (October 2024)","provider":"Anthropic","rawName":"Claude 3.5 Sonnet (October 2024)","variant":"compiled","setting":"Compiled by Epoch AI; effort: Oct 2024","score":4.08,"unit":"%","rank":49,"rankComputedBySite":true,"date":"2024-10-22","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"amazon-nova-lite","modelName":"Amazon Nova Lite","provider":"Amazon","rawName":"Amazon Nova Lite","variant":"compiled","setting":"Compiled by Epoch AI","score":3.64,"unit":"%","rank":50,"rankComputedBySite":true,"date":"2024-12-03","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"},{"model":"gpt-4o-nov-2024","modelName":"GPT-4o (Nov 2024)","provider":"OpenAI","rawName":"GPT-4o (Nov 2024)","variant":"compiled","setting":"Compiled by Epoch AI; effort: Nov 2024","score":2.72,"unit":"%","rank":51,"rankComputedBySite":true,"date":"2024-11-20","sourceType":"compiled","source":"epoch","sourceUrl":"https://epoch.ai/benchmarks","collected":"2026-10-10"}]}