Skip to content
Model Index
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
LabsMethodology
  1. Benchmarks
  2. Reasoning & science

Humanity's Last Exam

Independent testArtificial Analysis · updated Oct 8, 2026

What it measures

2,158 expert-written questions at the frontier of academic knowledge across maths, sciences and humanities.

How to read it

Designed to stay hard. A model above 50% is answering questions most specialists outside the field could not.

Open at Artificial Analysis
Best
61.4%
Claude Opus 5.5
Median
39.2%
DeepSeek V4.1 Flash

Results

39 of 41 tracked models have a result.

Filter in explorer
  1. 1
    Claude Opus 5.5

    Anthropic · Max effort

    61.4%
  2. 2
    Claude Fable 5.1

    Anthropic · Max effort

    59.1%
  3. 3
    Gemini 4 ArgonLimited

    Google DeepMind · High effort

    57.1%
  4. 4
    Claude Sonnet 5.5

    Anthropic · Max effort

    55.0%
  5. 5
    GPT-6 Astra

    OpenAI · Max effort

    54.7%
  6. 6
    GPT-6.1 Sol

    OpenAI · Max effort

    52.9%
  7. 7
    MiMo-V2.6-ProOpen

    Xiaomi

    49.4%
  8. 8
    Muse Spark 1.3

    Meta · Max effort

    48.7%
  9. 9
    Gemini 3.8 Flash

    Google DeepMind · High effort

    47.8%
  10. 10
    Gemini 3.1 Pro

    Google DeepMind

    47.0%
  11. 11
    Kimi K3Open

    Moonshot AI · Max effort

    46.9%
  12. 12
    Step 5 Preview

    StepFun

    46.5%
  13. 13
    Claude Haiku 5.5

    Anthropic · Max effort

    44.4%
  14. 14
    Grok 4.7

    SpaceXAI · xHigh effort

    43.1%
  15. 14
    Qwen3.8-Max

    Alibaba

    43.1%
  16. 16
    GPT-5.6 Terra

    OpenAI · Max effort

    42.9%
  17. 17
    GLM-5.3Open

    Z.ai · Max effort

    42.3%
  18. 18
    DeepSeek V4 ProOpen

    DeepSeek · Max effort

    41.0%
  19. 19
    GLM-5.3-FlashOpen

    Z.ai

    39.9%
  20. 20
    DeepSeek V4.1 FlashOpen

    DeepSeek · Max effort

    39.2%
  21. 21
    MiniMax-M3Open

    MiniMax

    39.0%
  22. 22
    GPT-6 Luna

    OpenAI · Max effort

    38.5%
  23. 23
    Qwen3.8-FlashOpen

    Alibaba

    38.0%
  24. 24
    MiMo-V2.6-FlashOpen

    Xiaomi

    35.1%
  25. 25
    Kimi K2.7 CodeOpen

    Moonshot AI

    35.0%
  26. 25
    Mistral Large 4

    Mistral AI

    35.0%
  27. 27
    Qwen3.8-27BOpen

    Alibaba · xHigh effort

    33.9%
  28. 28
    Hy3Open

    Tencent

    33.5%
  29. 29
    TM
    Inkling-SmallOpen

    Thinking Machines Lab

    33.3%
  30. 30
    TM
    InklingOpen

    Thinking Machines Lab · xHigh effort

    31.9%
  31. 31
    Nemotron 3 UltraOpen

    NVIDIA · Reasoning effort

    28.4%
  32. 32
    Gemma 4 31BOpen

    Google DeepMind · Reasoning effort

    23.6%
  33. 33
    Muse Glimmer 30BOpen

    Meta · High effort

    22.0%
  34. 34
    gpt-oss-120bOpen

    OpenAI · High effort

    19.6%
  35. 35
    Gemini 3.5 Flash-Lite

    Google DeepMind

    18.8%
  36. 36
    Mistral Medium 3.5Open

    Mistral AI

    13.8%
  37. 37
    Command A+Open

    Cohere

    12.0%
  38. 38
    Nova 2 Lite

    Amazon · High effort

    11.6%
  39. 39
    Mistral Small 4Open

    Mistral AI · Reasoning effort

    9.9%
2 tracked models without a result

Hy4 preview, MAI-Thinking-1.

Missing means the evaluator has not published a run, not that the model would score poorly.

Model Index

An independent snapshot of frontier SI models and how they perform on third-party benchmarks.

Data as of Oct 8, 2026

Explore

  • All models
  • Compare
  • Benchmarks
  • Model finder
  • Labs
  • Methodology & sources

Sources

Benchmark results come from Artificial Analysis, Arena, Vals AI and ARC Prize. Specifications and prices come from each lab's own documentation.

Not affiliated with any lab or evaluator. Names and logos are trademarks of their respective owners and are used only to identify them. Scores change often — check the linked source before making a decision that depends on one.

  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder