Skip to content

Humanity's Last Exam without tools

Tool-free variant of Humanity's Last Exam that isolates a model's raw frontier reasoning.

43 models

Higher is better.

  1. 1Claude Opus 5.5Anthropic64.4
  2. 2Claude Fable 5.1Anthropic60.9
  3. 3Claude Mythos 5Anthropic59.0
  4. 4Claude Sonnet 5.5Anthropic56.9
  5. 5Claude Opus 5Anthropic56.3
  6. 6Muse Spark 1.1Meta52.2
  7. 7Claude Opus 4.8Anthropic49.8
  8. 8Pareto 26.9Unbiased49.0
  9. 9Claude Opus 4.7 (Adaptive)Anthropic46.9
  10. 10Claude Haiku 5.5Anthropic45.9
  11. 11Gemini 3.1 ProGoogle45.4
  12. 12Ornith-1.5-397BOrnith AI44.6
  13. 13Qwen3.8 MaxAlibaba43.6
  14. 14Kimi K3Moonshot AI43.5
  15. 15MiniMax M3.1 FlashMiniMax43.4
  16. 16Hy4 previewTencent43.4
  17. 17Claude Sonnet 5Anthropic43.2
  18. 18GPT-5.5 ProOpenAI43.1
  19. 19Muse SparkMeta42.8
  20. 20GPT-5.4 ProOpenAI42.7
  21. 21GPT-5.5OpenAI41.4
  22. 22GLM-5.2Z.AI40.5
  23. 23Claude Opus 4.6Anthropic40.0
  24. 24GPT-5.4OpenAI39.8
  25. 25Beam 1.0 501B-A23BReflection AI36.0
  26. 26Qwen3.8-Flash-NextAlibaba35.9
  27. 27MiMo-V2.5-ProXiaomi34.0
  28. 28Nemotron 3.5 Super VL 120B A12BNVIDIA31.8
  29. 29Grok 4.20xAI31.6
  30. 30Inkling-SmallThinking Machines Lab31.6
  31. 31Qwen3.8-27BAlibaba30.8
  32. 32InklingThinking Machines Lab30.0
  33. 33Solar Open 2Upstage28.8
  34. 34GPT-5.4 miniOpenAI28.2
  35. 35Nemotron 3 UltraNVIDIA26.7
  36. 36Ornith-1.5-35B-A3BOrnith AI25.6
  37. 37Nemotron 3.5 Bolt 74B-A9BNVIDIA25.3
  38. 38GPT-5.4 nanoOpenAI24.3
  39. 39Ornith-1.5-9BOrnith AI20.2
  40. 40Gemma 4 31BGoogle19.5
Higher is better. Showing the top 40 of 43. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Tool-free expert QA
Tasks
Expert-level questions
Authors
OpenAI (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.

All results

1Claude Opus 5.5 Anthropic64.4
2Claude Fable 5.1 Anthropic60.9
3Claude Mythos 5 Anthropic59.0
4Claude Sonnet 5.5 Anthropic56.9
5Claude Opus 5 Anthropic56.3
6Muse Spark 1.1 Meta52.2
7Claude Opus 4.8 Anthropic49.8
8Pareto 26.9 Unbiased49.0
9Claude Opus 4.7 (Adaptive) Anthropic46.9
10Claude Haiku 5.5 Anthropic45.9
11Gemini 3.1 Pro Google45.4
12Ornith-1.5-397B Ornith AI44.6
13Qwen3.8 Max Alibaba43.6
14Kimi K3 Moonshot AI43.5
15MiniMax M3.1 Flash MiniMax43.4
16Hy4 preview Tencent43.4
17Claude Sonnet 5 Anthropic43.2
18GPT-5.5 Pro OpenAI43.1
19Muse Spark Meta42.8
20GPT-5.4 Pro OpenAI42.7
21GPT-5.5 OpenAI41.4
22GLM-5.2 Z.AI40.5
23Claude Opus 4.6 Anthropic40.0
24GPT-5.4 OpenAI39.8
25Beam 1.0 501B-A23B Reflection AI36.0
26Qwen3.8-Flash-Next Alibaba35.9
27MiMo-V2.5-Pro Xiaomi34.0
28Nemotron 3.5 Super VL 120B A12B NVIDIA31.8
29Grok 4.20 xAI31.6
30Inkling-Small Thinking Machines Lab31.6
31Qwen3.8-27B Alibaba30.8
32Inkling Thinking Machines Lab30.0
33Solar Open 2 Upstage28.8
34GPT-5.4 mini OpenAI28.2
35Nemotron 3 Ultra NVIDIA26.7
36Ornith-1.5-35B-A3B Ornith AI25.6
37Nemotron 3.5 Bolt 74B-A9B NVIDIA25.3
38GPT-5.4 nano OpenAI24.3
39Ornith-1.5-9B Ornith AI20.2
40Gemma 4 31B Google19.5
41Nemotron 3.5 Lightning 30B A3B NVFP4 NVIDIA10.5
42Gemma 4 26B A4B Google8.7
43Gemma 4 12B Google5.2