Skip to content

FrontierMath v2 Tiers 1-3

Epoch AI's corrected v2 core FrontierMath suite of private advanced mathematics problems. Models can reason iteratively and use Python; scores are pass rates on the private set.

64 models

Higher is better.

  1. 1GPT-6.1 SolOpenAI93.684
  2. 2GPT-6 AstraOpenAI93.684
  3. 3Claude Opus 5.5Anthropic91.228
  4. 4Claude Fable 5.1Anthropic90.175
  5. 5GPT-6 SolOpenAI89.825
  6. 6GPT-5.6 SolOpenAI89.000
  7. 7Claude Sonnet 5.5Anthropic88.772
  8. 8GPT-5.5 ProOpenAI87.719
  9. 9Claude Fable 5Anthropic87.018
  10. 10Gemini 4 ArgonGoogle85.965
  11. 11Claude Opus 5Anthropic85.614
  12. 12GPT-5.5OpenAI85.263
  13. 13GPT-5.6 TerraOpenAI84.900
  14. 14GPT-5.4 ProOpenAI82.456
  15. 15Claude Opus 4.8Anthropic80.000
  16. 16GPT-6 LunaOpenAI78.947
  17. 17GPT-5.6 LunaOpenAI78.600
  18. 18GPT-5.4OpenAI78.596
  19. 19Claude Haiku 5.5Anthropic75.088
  20. 20Qwen3.8 MaxAlibaba74.737
  21. 21Muse Spark 1.3Meta74.035
  22. 22Gemini 3.7 FlashGoogle71.579
  23. 23Claude Opus 4.7Anthropic70.175
  24. 24GLM-5.3Z.AI68.772
  25. 25Gemini 3.8 FlashGoogle68.421
  26. 26GPT-5.2OpenAI67.400
  27. 27DeepSeek V4.1 Flash 1017DeepSeek67.368
  28. 28Grok 4.6xAI65.965
  29. 29Claude Opus 4.6Anthropic65.965
  30. 30Claude Sonnet 5Anthropic65.614
  31. 31Qwen3.7 MaxAlibaba64.561
  32. 32Gemini 3.5 FlashGoogle62.807
  33. 33Gemini 3.1 ProGoogle59.649
  34. 34GLM-5.2Z.AI59.206
  35. 35Gemini 3.6 FlashGoogle58.947
  36. 36Grok 4.5xAI57.193
  37. 37Kimi K2.6Moonshot AI57.193
  38. 38GLM-5.3-FlashZ.AI55.789
  39. 39Kimi K2.7 CodeMoonshot AI54.035
  40. 40Grok 4.7xAI52.982
Higher is better. Showing the top 40 of 64. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Python-enabled iterative mathematical problem solving
Tasks
285 private advanced mathematics problems
Authors
Epoch AI (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

1GPT-6.1 Sol OpenAI93.684
2GPT-6 Astra OpenAI93.684
3Claude Opus 5.5 Anthropic91.228
4Claude Fable 5.1 Anthropic90.175
5GPT-6 Sol OpenAI89.825
6GPT-5.6 Sol OpenAI89.000
7Claude Sonnet 5.5 Anthropic88.772
8GPT-5.5 Pro OpenAI87.719
9Claude Fable 5 Anthropic87.018
10Gemini 4 Argon Google85.965
11Claude Opus 5 Anthropic85.614
12GPT-5.5 OpenAI85.263
13GPT-5.6 Terra OpenAI84.900
14GPT-5.4 Pro OpenAI82.456
15Claude Opus 4.8 Anthropic80.000
16GPT-6 Luna OpenAI78.947
17GPT-5.6 Luna OpenAI78.600
18GPT-5.4 OpenAI78.596
19Claude Haiku 5.5 Anthropic75.088
20Qwen3.8 Max Alibaba74.737
21Muse Spark 1.3 Meta74.035
22Gemini 3.7 Flash Google71.579
23Claude Opus 4.7 Anthropic70.175
24GLM-5.3 Z.AI68.772
25Gemini 3.8 Flash Google68.421
26GPT-5.2 OpenAI67.400
27DeepSeek V4.1 Flash 1017 DeepSeek67.368
28Grok 4.6 xAI65.965
29Claude Opus 4.6 Anthropic65.965
30Claude Sonnet 5 Anthropic65.614
31Qwen3.7 Max Alibaba64.561
32Gemini 3.5 Flash Google62.807
33Gemini 3.1 Pro Google59.649
34GLM-5.2 Z.AI59.206
35Gemini 3.6 Flash Google58.947
36Grok 4.5 xAI57.193
37Kimi K2.6 Moonshot AI57.193
38GLM-5.3-Flash Z.AI55.789
39Kimi K2.7 Code Moonshot AI54.035
40Grok 4.7 xAI52.982
41Gemini 3 Flash Google51.228
42GPT-5.4 mini OpenAI51.228
43Inkling-Small Thinking Machines Lab46.316
44GPT-5.4 nano OpenAI44.912
45Grok 4.3 xAI42.807
46Qwen3.6 Plus Alibaba38.246
47GLM-5.1 Z.AI36.842
48o4-mini (high) OpenAI36.140
49Qwen3.6-27B Alibaba35.088
50Claude Opus 4.5 Anthropic34.386
51Qwen3.7 Plus Alibaba34.386
52Inkling Thinking Machines Lab33.333
53o3 OpenAI33.333
54Gemini 3.1 Flash-Lite Google27.719
55Gemini 3.5 Flash-Lite Google25.965
56Gemini 2.5 Pro Google24.561
57Claude Sonnet 4.5 Anthropic23.860
58Qwen3.6-35B-A3B Alibaba20.351
59Qwen3.7 Flash Alibaba19.298
60Qwen3.5 Flash Alibaba18.246
61o1 OpenAI14.737
62GPT-4.1 mini OpenAI6.667
63GPT-4.1 OpenAI5.965
64GPT-4o OpenAI0.351