Benchmark Almanac

FrontierMath

gpt-5.5-pro-pre-release 52.4% tops its leaderboard GPT-5.4 Pro 50% leads the models we track
GPT-5.5 Pro 52.4% leads since April 2026
GPT-5.5 Pro 52.4% leads since April 2026
Fading

Epoch AI’s run of GPT-5.5 Pro, released April 2026, scored 52.4%. The official leaderboard hasn’t tested it yet.

Unpublished research-level math problems with automatically verifiable answers. Tier 4 is the hardest tier.

Problems held privately by Epoch AI.

Made by
Epoch AI
Released
November 2024
Kind
Research mathematics
Level
Expert
Scored
Checked automatically
Answers
Short answer
Tasks
Kept private
Strongest model tested
GPT-5.5 Pro #14 on the capability index
Last in a launch post
June 2026, dropped for a newer version
Last model added
June 2026
Scores refreshed
November 12th, 2026

In the authors’ words

We introduce FrontierMath, a benchmark of hundreds of original, exceptionally challenging mathematics problems crafted and vetted by expert mathematicians. The questions cover most major branches of modern mathematics – from computationally intensive problems in number theory and real analysis to abstract questions in algebraic geometry and category theory. Solving a typical problem requires multiple hours of effort from a researcher in the relevant branch of mathematics, and for the upper end questions, multiple days. FrontierMath uses new, unpublished problems and automated verification to reliably evaluate models while minimizing risk of data contamination. Current state-of-the-art AI models solve under 2% of problems, revealing a vast gap between AI capabilities and the prowess of the mathematical community. As AI systems advance toward expert-level mathematical abilities, FrontierMath offers a rigorous testbed that quantifies their progress.

Abstract of FrontierMath: A Benchmark for Evaluating Advanced Mathematical Reasoning in AI (arXiv:2411.04872, CC0)
69
Entries on its own leaderboard, 55 models we track
12
Launch posts list it, from 2 labs
1
System card reports it, from 1 lab

gpt-5.5-pro-pre-release tops the leaderboard with 52.4%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is GPT-5.4 Pro, at 50%.

GPT-5.5 Pro leads with 52.4% across all independent runs

Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Each bar and row names its source.

GPT-5.5 Pro leads with 52.4%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

The newest model with a score here came out in May 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

69 entries. Scroll sideways for the rest.

  1. 52.4 gpt-5.5-pro-pre-release
  2. 51.7 gpt-5.5-pre-release
  3. 50.0 GPT-5.4 Pro
  4. 47.6 GPT-5.4
  5. 47.2 Claude Opus 4.8
  6. 43.8 Claude Opus 4.7
  7. 40.7 GPT-5.2
  8. 40.7 Claude Opus 4.6
  9. 39.0 Muse Spark
  10. 39.0 Kimi K2.6
  11. 39.0 Gemini 3.5 Flash
  12. 37.6 Gemini 3 Pro
  13. 36.9 Gemini 3.1 Pro
  14. 35.6 Gemini 3 Flash
  15. 33.5 GLM-5.1
  16. 32.4 GPT-5
  17. 32.4 Claude Sonnet 4.6
  18. 31.0 GPT-5.1
  19. 29.0 gemini-2.5-deep-think-2025-08-01-webapp
  20. 28.3 GPT-5.4 Mini
  21. 27.9 fireworks/kimi-k2p5
  22. 27.2 GPT-5 mini
  23. 26.2 Qwen 3.6 Plus
  24. 25.9 GPT-5.4 Nano
  25. 24.8 o4-mini
  26. 23.1 Qwen 3.6 Max (Preview)
  27. 22.1 fireworks/deepseek-v3p2
  28. 21.4 Kimi K2 Thinking
  29. 21.0 Qwen3.5-Plus
  30. 20.7 Claude Opus 4.5
  31. 19.7 grok-4-0709
  32. 18.7 o3
  33. 16.4 GLM-5
  34. 15.2 Claude Sonnet 4.5
  35. 14.1 Gemini 2.5 Pro
  36. 12.4 o3-mini
  37. 10.3 Qwen 3.6 Flash
  38. 9.31 o1
  39. 8.48 Qwen3-235B-A22B-Thinking
  40. 8.28 GPT-5 nano
  41. 7.24 Claude Opus 4.1
  42. 6.21 Qwen3.5-Flash
  43. 5.90 Claude Haiku 4.5
  44. 5.86 Grok-3 mini
  45. 5.52 GPT-4.1
  46. 4.84 Gemini 2.5 Flash
  47. 4.48 GPT-4.1 mini
  48. 4.48 Claude Opus 4
  49. 4.14 Claude 3.7 Sonnet
  50. 4.14 Claude Sonnet 4
  51. 3.82 GLM-4.6
  52. 3.79 Grok 3
  53. 2.44 GLM-4.7
  54. 2.07 Claude 3.5 Sonnet
  55. 1.72 o1-mini
  56. 1.72 DeepSeek-V3
  57. 1.72 gemini-2.0-flash-001
  58. 1.72 Qwen Plus
  59. 1.03 qwen-max-2025-01-25
  60. 1.03 GPT-4.1 nano
  61. 0.69 grok-2-1212
  62. 0.69 Llama-4-Maverick-17B-128E-Instruct-FP8
  63. 0.35 mistral-medium-2505
  64. 0.34 GPT-4o
  65. 0.34 Claude 3.5 Haiku
  66. 0.34 GPT-4o
  67. 0.34 mistral-large-2411
  68. 0.0 gemini-1.5-flash-002
  69. 0.0 Llama-4-Scout-17B-16E-Instruct
  1. 1 gpt-5.5-pro-pre-release high 52.4%
  2. 2 gpt-5.5-pre-release xhigh 51.7%
  3. 3 GPT-5.4 Pro xhigh; task version 1.1.5 50%
  4. 4 GPT-5.4 xhigh; task version 1.1.5 47.6%
  5. 5 Claude Opus 4.8 max; task version 1.1.9 47.2%
  6. 6 Claude Opus 4.7 xhigh; task version 1.1.6 43.8%
  7. 7 GPT-5.2 xhigh; task version 1.1.2 40.7%
  8. 8 Claude Opus 4.6 max; task version 1.1.5 40.7%
  9. 9 Muse Spark 39%
  10. 10 Kimi K2.6 task version 1.1.8 39%
  11. 11 Gemini 3.5 Flash high; task version 1.1.8 39%
  12. 12 Gemini 3 Pro task version 1.1.1 37.6%
  13. 13 Gemini 3.1 Pro task version 1.1.5 36.9%
  14. 14 Gemini 3 Flash task version 1.1.2 35.6%
  15. 15 GLM-5.1 task version 1.1.8 33.5%
  16. 16 GPT-5 high; task version 1.1.0 32.4%
  17. 17 Claude Sonnet 4.6 16K; task version 1.1.5 32.4%
  18. 18 GPT-5.1 high; task version 1.1.0 31%
  19. 19 gemini-2.5-deep-think-2025-08-01-webapp task version 1.0.0 29%
  20. 20 GPT-5.4 Mini high; task version 1.1.5 28.3%
  21. 21 fireworks/kimi-k2p5 task version 1.1.4 27.9%
  22. 22 GPT-5 mini high; task version 1.1.0 27.2%
  23. 23 Qwen 3.6 Plus task version 1.1.8 26.2%
  24. 24 GPT-5.4 Nano high; task version 1.1.5 25.9%
  25. 25 o4-mini high; task version 1.1.0 24.8%
  26. 26 Qwen 3.6 Max (Preview) task version 1.1.8 23.1%
  27. 27 fireworks/deepseek-v3p2 task version 1.1.2 22.1%
  28. 28 Kimi K2 Thinking task version 1.1.1 21.4%
  29. 29 Qwen3.5-Plus task version 1.1.8 21%
  30. 30 Claude Opus 4.5 task version 1.1.1 20.7%
  31. 31 grok-4-0709 task version 1.1.0 19.7%
  32. 32 o3 high; task version 1.1.0 18.7%
  33. 33 GLM-5 task version 1.1.5 16.4%
  34. 34 Claude Sonnet 4.5 32K; task version 1.1.0 15.2%
  35. 35 Gemini 2.5 Pro Jun 2025, task version 1.1.1 14.1%
  36. 36 o3-mini high; task version 1.1.0 12.4%
  37. 37 Qwen 3.6 Flash task version 1.1.8 10.3%
  38. 38 o1 high; task version 1.0.0 9.31%
  39. 39 Qwen3-235B-A22B-Thinking Jul 2025, task version 1.1.2 8.48%
  40. 40 GPT-5 nano high; task version 1.0.1 8.28%
  41. 41 Claude Opus 4.1 27K; task version 1.0.0 7.24%
  42. 42 Qwen3.5-Flash task version 1.1.8 6.21%
  43. 43 Claude Haiku 4.5 32K; task version 1.0.0 5.90%
  44. 44 Grok-3 mini high; task version 1.0.0 5.86%
  45. 45 GPT-4.1 task version 1.0.0 5.52%
  46. 46 Gemini 2.5 Flash Sep 2025, task version 1.1.2 4.84%
  47. 47 GPT-4.1 mini task version 1.0.0 4.48%
  48. 48 Claude Opus 4 task version 1.0.0 4.48%
  49. 49 Claude 3.7 Sonnet 16K; task version 1.0.0 4.14%
  50. 50 Claude Sonnet 4 task version 1.0.0 4.14%
  51. 51 GLM-4.6 task version 1.1.1 3.82%
  52. 52 Grok 3 task version 1.0.0 3.79%
  53. 53 GLM-4.7 task version 1.1.4 2.44%
  54. 54 Claude 3.5 Sonnet task version 1.0.0 2.07%
  55. 55 o1-mini medium; task version 1.0.0 1.72%
  56. 56 DeepSeek-V3 task version 1.0.0 1.72%
  57. 57 gemini-2.0-flash-001 task version 1.0.0 1.72%
  58. 58 Qwen Plus Apr 2025, task version 1.0.0 1.72%
  59. 59 qwen-max-2025-01-25 task version 1.0.0 1.03%
  60. 60 GPT-4.1 nano task version 1.0.0 1.03%
  61. 61 grok-2-1212 task version 1.0.0 0.69%
  62. 62 Llama-4-Maverick-17B-128E-Instruct-FP8 task version 1.0.0 0.69%
  63. 63 mistral-medium-2505 task version 1.0.0 0.35%
  64. 64 GPT-4o Aug 2024, task version 1.0.0 0.34%
  65. 65 Claude 3.5 Haiku task version 1.0.0 0.34%
  66. 66 GPT-4o Nov 2024, task version 1.0.0 0.34%
  67. 67 mistral-large-2411 task version 1.0.0 0.34%
  68. 68 gemini-1.5-flash-002 task version 1.0.0 0%
  69. 69 Llama-4-Scout-17B-16E-Instruct task version 1.0.0 0%

Source: Official FrontierMath leaderboard

73 models. Scroll sideways for the rest.

  1. 52.4 GPT-5.5 Pro
  2. 51.7 GPT-5.5
  3. 50.0 GPT-5.4 Pro
  4. 47.6 GPT-5.4
  5. 47.2 Claude Opus 4.8
  6. 43.8 Claude Opus 4.7
  7. 40.7 GPT-5.2
  8. 40.7 Claude Opus 4.6
  9. 39.0 Muse Spark
  10. 39.0 Kimi K2.6
  11. 39.0 Gemini 3.5 Flash
  12. 37.6 Gemini 3 Pro
  13. 36.9 Gemini 3.1 Pro
  14. 35.6 Gemini 3 Flash
  15. 33.5 GLM-5.1
  16. 32.4 GPT-5
  17. 32.4 Claude Sonnet 4.6
  18. 31.0 GPT-5.1
  19. 29.0 Gemini 2.5 Deep Think
  20. 28.3 GPT-5.4 Mini
  21. 27.9 Kimi K2.5
  22. 27.2 GPT-5 mini
  23. 26.2 Qwen 3.6 Plus
  24. 25.9 GPT-5.4 Nano
  25. 24.8 o4-mini
  26. 23.1 Qwen 3.6 Max (Preview)
  27. 22.1 DeepSeek-V3.2
  28. 21.4 Kimi K2 Thinking
  29. 21.0 Qwen3.5-Plus
  30. 21.0 Qwen 3.5 Plus (hosted 397B-A17B)
  31. 20.7 Claude Opus 4.5
  32. 19.7 Grok 4
  33. 18.7 o3
  34. 16.4 GLM-5
  35. 15.2 Claude Sonnet 4.5
  36. 14.1 Gemini 2.5 Pro
  37. 12.4 o3-mini
  38. 10.3 Qwen 3.6 Flash
  39. 9.31 o1
  40. 8.48 Qwen3-235B-A22B-Thinking
  41. 8.28 GPT-5 nano
  42. 7.24 Claude Opus 4.1
  43. 6.21 Qwen 3.5 Flash (hosted 35B-A3B)
  44. 6.21 Qwen3.5-Flash
  45. 5.90 Claude Haiku 4.5
  46. 5.86 Grok-3 mini
  47. 5.52 GPT-4.1
  48. 4.84 Gemini 2.5 Flash
  49. 4.84 Gemini 2.5 Flash
  50. 4.48 GPT-4.1 mini
  51. 4.48 Claude Opus 4
  52. 4.14 Claude 3.7 Sonnet
  53. 4.14 Claude Sonnet 4
  54. 3.82 GLM-4.6
  55. 3.79 Grok 3
  56. 2.44 GLM-4.7
  57. 2.07 Claude 3.5 Sonnet (October 2024)
  58. 2.07 Claude 3.5 Sonnet
  59. 1.72 o1-mini
  60. 1.72 DeepSeek-V3
  61. 1.72 Qwen Plus
  62. 1.72 Gemini 2.0 Flash
  63. 1.03 GPT-4.1 nano
  64. 1.03 Qwen2.5-Max
  65. 0.69 Grok-2
  66. 0.69 Llama 4 Maverick
  67. 0.35 Mistral Medium 3
  68. 0.34 GPT-4o
  69. 0.34 Claude 3.5 Haiku
  70. 0.34 GPT-4o
  71. 0.34 Mistral Large 2
  72. 0.0 Gemini 1.5 Flash
  73. 0.0 Llama 4 Scout
  1. 1 GPT-5.5 Pro high 52.4%
  2. 2 GPT-5.5 xhigh 51.7%
  3. 3 GPT-5.4 Pro xhigh; task version 1.1.5 50%
  4. 4 GPT-5.4 xhigh; task version 1.1.5 47.6%
  5. 5 Claude Opus 4.8 max; task version 1.1.9 47.2%
  6. 6 Claude Opus 4.7 xhigh; task version 1.1.6 43.8%
  7. 7 GPT-5.2 xhigh; task version 1.1.2 40.7%
  8. 8 Claude Opus 4.6 max; task version 1.1.5 40.7%
  9. 9 Muse Spark 39%
  10. 10 Kimi K2.6 task version 1.1.8 39%
  11. 11 Gemini 3.5 Flash high; task version 1.1.8 39%
  12. 12 Gemini 3 Pro task version 1.1.1 37.6%
  13. 13 Gemini 3.1 Pro task version 1.1.5 36.9%
  14. 14 Gemini 3 Flash task version 1.1.2 35.6%
  15. 15 GLM-5.1 task version 1.1.8 33.5%
  16. 16 GPT-5 high; task version 1.1.0 32.4%
  17. 17 Claude Sonnet 4.6 16K; task version 1.1.5 32.4%
  18. 18 GPT-5.1 high; task version 1.1.0 31%
  19. 19 Gemini 2.5 Deep Think 29%
  20. 20 GPT-5.4 Mini high; task version 1.1.5 28.3%
  21. 21 Kimi K2.5 Fireworks 27.9%
  22. 22 GPT-5 mini high; task version 1.1.0 27.2%
  23. 23 Qwen 3.6 Plus task version 1.1.8 26.2%
  24. 24 GPT-5.4 Nano high; task version 1.1.5 25.9%
  25. 25 o4-mini high; task version 1.1.0 24.8%
  26. 26 Qwen 3.6 Max (Preview) task version 1.1.8 23.1%
  27. 27 DeepSeek-V3.2 Thinking; Fireworks 22.1%
  28. 28 Kimi K2 Thinking task version 1.1.1 21.4%
  29. 29 Qwen3.5-Plus task version 1.1.8 21%
  30. 30 Qwen 3.5 Plus (hosted 397B-A17B) 21%
  31. 31 Claude Opus 4.5 task version 1.1.1 20.7%
  32. 32 Grok 4 19.7%
  33. 33 o3 high; task version 1.1.0 18.7%
  34. 34 GLM-5 task version 1.1.5 16.4%
  35. 35 Claude Sonnet 4.5 32K; task version 1.1.0 15.2%
  36. 36 Gemini 2.5 Pro Jun 2025, task version 1.1.1 14.1%
  37. 37 o3-mini high; task version 1.1.0 12.4%
  38. 38 Qwen 3.6 Flash task version 1.1.8 10.3%
  39. 39 o1 high; task version 1.0.0 9.31%
  40. 40 Qwen3-235B-A22B-Thinking Jul 2025, task version 1.1.2 8.48%
  41. 41 GPT-5 nano high 8.28%
  42. 42 Claude Opus 4.1 27K; task version 1.0.0 7.24%
  43. 43 Qwen 3.5 Flash (hosted 35B-A3B) 6.21%
  44. 44 Qwen3.5-Flash task version 1.1.8 6.21%
  45. 45 Claude Haiku 4.5 32K; task version 1.0.0 5.90%
  46. 46 Grok-3 mini high; task version 1.0.0 5.86%
  47. 47 GPT-4.1 5.52%
  48. 48 Gemini 2.5 Flash Sep 2025, task version 1.1.2 4.84%
  49. 49 Gemini 2.5 Flash Jun 2025 4.84%
  50. 50 GPT-4.1 mini task version 1.0.0 4.48%
  51. 51 Claude Opus 4 task version 1.0.0 4.48%
  52. 52 Claude 3.7 Sonnet 16k thinking 4.14%
  53. 53 Claude Sonnet 4 4.14%
  54. 54 GLM-4.6 Together 3.82%
  55. 55 Grok 3 task version 1.0.0 3.79%
  56. 56 GLM-4.7 Together 2.44%
  57. 57 Claude 3.5 Sonnet (October 2024) 2.07%
  58. 58 Claude 3.5 Sonnet task version 1.0.0 2.07%
  59. 59 o1-mini medium; task version 1.0.0 1.72%
  60. 60 DeepSeek-V3 task version 1.0.0 1.72%
  61. 61 Qwen Plus Apr 2025, task version 1.0.0 1.72%
  62. 62 Gemini 2.0 Flash Feb 2025 1.72%
  63. 63 GPT-4.1 nano task version 1.0.0 1.03%
  64. 64 Qwen2.5-Max 1.03%
  65. 65 Grok-2 Dec 2024 0.69%
  66. 66 Llama 4 Maverick FP8 0.69%
  67. 67 Mistral Medium 3 0.35%
  68. 68 GPT-4o Aug 2024, task version 1.0.0 0.34%
  69. 69 Claude 3.5 Haiku task version 1.0.0 0.34%
  70. 70 GPT-4o Nov 2024, task version 1.0.0 0.34%
  71. 71 Mistral Large 2 Nov 2024 0.34%
  72. 72 Gemini 1.5 Flash Sep 2024 0%
  73. 73 Llama 4 Scout 0%

Sources: Epoch AI runs (CC BY 4.0) and official FrontierMath leaderboard

75 models. Scroll sideways for the rest.

  1. 52.4 GPT-5.5 Pro
  2. 52.4 GPT-5.5
  3. 50.0 GPT-5.4 Pro
  4. 47.6 GPT-5.4
  5. 47.2 Claude Opus 4.8
  6. 43.8 Claude Opus 4.7
  7. 40.7 GPT-5.2
  8. 40.7 Claude Opus 4.6
  9. 39.0 Muse Spark
  10. 39.0 Kimi K2.6
  11. 39.0 Gemini 3.5 Flash
  12. 37.6 Gemini 3 Pro
  13. 36.9 Gemini 3.1 Pro
  14. 35.6 Gemini 3 Flash
  15. 33.5 GLM-5.1
  16. 32.4 GPT-5
  17. 32.4 Claude Sonnet 4.6
  18. 31.0 GPT-5.1
  19. 29.0 Gemini 2.5 Deep Think
  20. 28.3 GPT-5.4 Mini
  21. 27.9 Kimi K2.5
  22. 27.4 ChatGPT agent
  23. 27.2 GPT-5 mini
  24. 26.2 Qwen 3.6 Plus
  25. 25.9 GPT-5.4 Nano
  26. 24.8 o4-mini
  27. 23.1 Qwen 3.6 Max (Preview)
  28. 22.1 DeepSeek-V3.2
  29. 21.4 Kimi K2 Thinking
  30. 21.0 Qwen3.5-Plus
  31. 21.0 Qwen 3.5 Plus (hosted 397B-A17B)
  32. 20.7 Claude Opus 4.5
  33. 19.7 Grok 4
  34. 18.7 o3
  35. 16.4 GLM-5
  36. 15.2 Claude Sonnet 4.5
  37. 14.1 Gemini 2.5 Pro
  38. 12.4 o3-mini
  39. 10.3 Qwen 3.6 Flash
  40. 9.60 GPT-5 nano
  41. 9.31 o1
  42. 8.48 Qwen3-235B-A22B-Thinking
  43. 7.24 Claude Opus 4.1
  44. 6.30 MAI-Code-1-Flash
  45. 6.21 Qwen 3.5 Flash (hosted 35B-A3B)
  46. 6.21 Qwen3.5-Flash
  47. 5.90 Claude Haiku 4.5
  48. 5.86 Grok-3 mini
  49. 5.52 GPT-4.1
  50. 4.84 Gemini 2.5 Flash
  51. 4.84 Gemini 2.5 Flash
  52. 4.48 GPT-4.1 mini
  53. 4.48 Claude Opus 4
  54. 4.14 Claude 3.7 Sonnet
  55. 4.14 Claude Sonnet 4
  56. 3.82 GLM-4.6
  57. 3.79 Grok 3
  58. 2.44 GLM-4.7
  59. 2.07 Claude 3.5 Sonnet (October 2024)
  60. 2.07 Claude 3.5 Sonnet
  61. 1.72 o1-mini
  62. 1.72 DeepSeek-V3
  63. 1.72 Qwen Plus
  64. 1.72 Gemini 2.0 Flash
  65. 1.03 GPT-4.1 nano
  66. 1.03 Qwen2.5-Max
  67. 0.69 Grok-2
  68. 0.69 Llama 4 Maverick
  69. 0.35 Mistral Medium 3
  70. 0.34 GPT-4o
  71. 0.34 Claude 3.5 Haiku
  72. 0.34 GPT-4o
  73. 0.34 Mistral Large 2
  74. 0.0 Gemini 1.5 Flash
  75. 0.0 Llama 4 Scout
  1. 1 GPT-5.5 Pro high 52.4%
  2. 2 GPT-5.5 52.4%
  3. 3 GPT-5.4 Pro xhigh; task version 1.1.5 50%
  4. 4 GPT-5.4 xhigh; task version 1.1.5 47.6%
  5. 5 Claude Opus 4.8 max; task version 1.1.9 47.2%
  6. 6 Claude Opus 4.7 xhigh; task version 1.1.6 43.8%
  7. 7 GPT-5.2 xhigh; task version 1.1.2 40.7%
  8. 8 Claude Opus 4.6 max; task version 1.1.5 40.7%
  9. 9 Muse Spark 39%
  10. 10 Kimi K2.6 task version 1.1.8 39%
  11. 11 Gemini 3.5 Flash high; task version 1.1.8 39%
  12. 12 Gemini 3 Pro task version 1.1.1 37.6%
  13. 13 Gemini 3.1 Pro task version 1.1.5 36.9%
  14. 14 Gemini 3 Flash task version 1.1.2 35.6%
  15. 15 GLM-5.1 task version 1.1.8 33.5%
  16. 16 GPT-5 high; task version 1.1.0 32.4%
  17. 17 Claude Sonnet 4.6 16K; task version 1.1.5 32.4%
  18. 18 GPT-5.1 high; task version 1.1.0 31%
  19. 19 Gemini 2.5 Deep Think 29%
  20. 20 GPT-5.4 Mini high; task version 1.1.5 28.3%
  21. 21 Kimi K2.5 Fireworks 27.9%
  22. 22 ChatGPT agent 27.4%
  23. 23 GPT-5 mini high; task version 1.1.0 27.2%
  24. 24 Qwen 3.6 Plus task version 1.1.8 26.2%
  25. 25 GPT-5.4 Nano high; task version 1.1.5 25.9%
  26. 26 o4-mini high; task version 1.1.0 24.8%
  27. 27 Qwen 3.6 Max (Preview) task version 1.1.8 23.1%
  28. 28 DeepSeek-V3.2 Thinking; Fireworks 22.1%
  29. 29 Kimi K2 Thinking task version 1.1.1 21.4%
  30. 30 Qwen3.5-Plus task version 1.1.8 21%
  31. 31 Qwen 3.5 Plus (hosted 397B-A17B) 21%
  32. 32 Claude Opus 4.5 task version 1.1.1 20.7%
  33. 33 Grok 4 19.7%
  34. 34 o3 high; task version 1.1.0 18.7%
  35. 35 GLM-5 task version 1.1.5 16.4%
  36. 36 Claude Sonnet 4.5 32K; task version 1.1.0 15.2%
  37. 37 Gemini 2.5 Pro Jun 2025, task version 1.1.1 14.1%
  38. 38 o3-mini high; task version 1.1.0 12.4%
  39. 39 Qwen 3.6 Flash task version 1.1.8 10.3%
  40. 40 GPT-5 nano 9.60%
  41. 41 o1 high; task version 1.0.0 9.31%
  42. 42 Qwen3-235B-A22B-Thinking Jul 2025, task version 1.1.2 8.48%
  43. 43 Claude Opus 4.1 27K; task version 1.0.0 7.24%
  44. 44 MAI-Code-1-Flash 6.30%
  45. 45 Qwen 3.5 Flash (hosted 35B-A3B) 6.21%
  46. 46 Qwen3.5-Flash task version 1.1.8 6.21%
  47. 47 Claude Haiku 4.5 32K; task version 1.0.0 5.90%
  48. 48 Grok-3 mini high; task version 1.0.0 5.86%
  49. 49 GPT-4.1 5.52%
  50. 50 Gemini 2.5 Flash Sep 2025, task version 1.1.2 4.84%
  51. 51 Gemini 2.5 Flash Jun 2025 4.84%
  52. 52 GPT-4.1 mini task version 1.0.0 4.48%
  53. 53 Claude Opus 4 task version 1.0.0 4.48%
  54. 54 Claude 3.7 Sonnet 16k thinking 4.14%
  55. 55 Claude Sonnet 4 4.14%
  56. 56 GLM-4.6 Together 3.82%
  57. 57 Grok 3 task version 1.0.0 3.79%
  58. 58 GLM-4.7 Together 2.44%
  59. 59 Claude 3.5 Sonnet (October 2024) 2.07%
  60. 60 Claude 3.5 Sonnet task version 1.0.0 2.07%
  61. 61 o1-mini medium; task version 1.0.0 1.72%
  62. 62 DeepSeek-V3 task version 1.0.0 1.72%
  63. 63 Qwen Plus Apr 2025, task version 1.0.0 1.72%
  64. 64 Gemini 2.0 Flash Feb 2025 1.72%
  65. 65 GPT-4.1 nano task version 1.0.0 1.03%
  66. 66 Qwen2.5-Max 1.03%
  67. 67 Grok-2 Dec 2024 0.69%
  68. 68 Llama 4 Maverick FP8 0.69%
  69. 69 Mistral Medium 3 0.35%
  70. 70 GPT-4o Aug 2024, task version 1.0.0 0.34%
  71. 71 Claude 3.5 Haiku task version 1.0.0 0.34%
  72. 72 GPT-4o Nov 2024, task version 1.0.0 0.34%
  73. 73 Mistral Large 2 Nov 2024 0.34%
  74. 74 Gemini 1.5 Flash Sep 2024 0%
  75. 75 Llama 4 Scout 0%

Sources: Epoch AI runs (CC BY 4.0), labs’ model cards and official FrontierMath leaderboard

#EntryScoreSettingDateSource
1 gpt-5.5-pro-pre-release 52.4% high, Epoch Apr 2026 Official leaderboard
2 gpt-5.5-pre-release 51.7% xhigh, Epoch Apr 2026 Official leaderboard
3 GPT-5.4 Pro 50% xhigh; task version 1.1.5, Epoch Mar 2026 Official leaderboard
4 GPT-5.4 47.6% xhigh; task version 1.1.5, Epoch Mar 2026 Official leaderboard
5 Claude Opus 4.8 47.2% max; task version 1.1.9, Epoch May 2026 Official leaderboard
6 Claude Opus 4.7 43.8% xhigh; task version 1.1.6, Epoch Apr 2026 Official leaderboard
7 GPT-5.2 40.7% xhigh; task version 1.1.2, Epoch Dec 2025 Official leaderboard
8 Claude Opus 4.6 40.7% max; task version 1.1.5, Epoch Feb 2026 Official leaderboard
9 Muse Spark 39% Epoch Apr 2026 Official leaderboard
10 Kimi K2.6 39% task version 1.1.8, Epoch Apr 2026 Official leaderboard
11 Gemini 3.5 Flash 39% high; task version 1.1.8, Epoch May 2026 Official leaderboard
12 Gemini 3 Pro 37.6% task version 1.1.1, Epoch Nov 2025 Official leaderboard
13 Gemini 3.1 Pro 36.9% task version 1.1.5, Epoch Feb 2026 Official leaderboard
14 Gemini 3 Flash 35.6% task version 1.1.2, Epoch Dec 2025 Official leaderboard
15 GLM-5.1 33.5% task version 1.1.8, Epoch Apr 2026 Official leaderboard
16 GPT-5 32.4% high; task version 1.1.0, Epoch Aug 2025 Official leaderboard
17 Claude Sonnet 4.6 32.4% 16K; task version 1.1.5, Epoch Feb 2026 Official leaderboard
18 GPT-5.1 31% high; task version 1.1.0, Epoch Nov 2025 Official leaderboard
19 gemini-2.5-deep-think-2025-08-01-webapp 29% task version 1.0.0, Epoch Official leaderboard
20 GPT-5.4 Mini 28.3% high; task version 1.1.5, Epoch Mar 2026 Official leaderboard
21 fireworks/kimi-k2p5 27.9% task version 1.1.4, Epoch Feb 2026 Official leaderboard
22 GPT-5 mini 27.2% high; task version 1.1.0, Epoch Aug 2025 Official leaderboard
23 Qwen 3.6 Plus 26.2% task version 1.1.8, Epoch Mar 2026 Official leaderboard
24 GPT-5.4 Nano 25.9% high; task version 1.1.5, Epoch Mar 2026 Official leaderboard
25 o4-mini 24.8% high; task version 1.1.0, Epoch Apr 2025 Official leaderboard
26 Qwen 3.6 Max (Preview) 23.1% task version 1.1.8, Epoch Apr 2026 Official leaderboard
27 fireworks/deepseek-v3p2 22.1% task version 1.1.2, Epoch Dec 2025 Official leaderboard
28 Kimi K2 Thinking 21.4% task version 1.1.1, Epoch Nov 2025 Official leaderboard
29 Qwen3.5-Plus 21% task version 1.1.8, Epoch Feb 2026 Official leaderboard
30 Claude Opus 4.5 20.7% task version 1.1.1, Epoch Nov 2025 Official leaderboard
31 grok-4-0709 19.7% task version 1.1.0, Epoch Nov 2025 Official leaderboard
32 o3 18.7% high; task version 1.1.0, Epoch Apr 2025 Official leaderboard
33 GLM-5 16.4% task version 1.1.5, Epoch Feb 2026 Official leaderboard
34 Claude Sonnet 4.5 15.2% 32K; task version 1.1.0, Epoch Sep 2025 Official leaderboard
35 Gemini 2.5 Pro Jun 2025 14.1% task version 1.1.1, Epoch Jun 2025 Official leaderboard
36 o3-mini 12.4% high; task version 1.1.0, Epoch Jan 2025 Official leaderboard
37 Qwen 3.6 Flash 10.3% task version 1.1.8, Epoch Apr 2026 Official leaderboard
38 o1 9.31% high; task version 1.0.0, Epoch Dec 2024 Official leaderboard
39 Qwen3-235B-A22B-Thinking Jul 2025 8.48% task version 1.1.2, Epoch Jul 2025 Official leaderboard
40 GPT-5 nano 8.28% high; task version 1.0.1, Epoch Aug 2025 Official leaderboard
41 Claude Opus 4.1 7.24% 27K; task version 1.0.0, Epoch Aug 2025 Official leaderboard
42 Qwen3.5-Flash 6.21% task version 1.1.8, Epoch Feb 2026 Official leaderboard
43 Claude Haiku 4.5 5.90% 32K; task version 1.0.0, Epoch Oct 2025 Official leaderboard
44 Grok-3 mini 5.86% high; task version 1.0.0, Epoch Apr 2025 Official leaderboard
45 GPT-4.1 5.52% task version 1.0.0, Epoch Apr 2025 Official leaderboard
46 Gemini 2.5 Flash Sep 2025 4.84% task version 1.1.2, Epoch Sep 2025 Official leaderboard
47 GPT-4.1 mini 4.48% task version 1.0.0, Epoch Apr 2025 Official leaderboard
48 Claude Opus 4 4.48% task version 1.0.0, Epoch May 2025 Official leaderboard
49 Claude 3.7 Sonnet 4.14% 16K; task version 1.0.0, Epoch Feb 2025 Official leaderboard
50 Claude Sonnet 4 4.14% task version 1.0.0, Epoch May 2025 Official leaderboard
51 GLM-4.6 3.82% task version 1.1.1, Epoch Sep 2025 Official leaderboard
52 Grok 3 3.79% task version 1.0.0, Epoch Apr 2025 Official leaderboard
53 GLM-4.7 2.44% task version 1.1.4, Epoch Dec 2025 Official leaderboard
54 Claude 3.5 Sonnet 2.07% task version 1.0.0, Epoch Jun 2024 Official leaderboard
55 o1-mini 1.72% medium; task version 1.0.0, Epoch Sep 2024 Official leaderboard
56 DeepSeek-V3 1.72% task version 1.0.0, Epoch Dec 2024 Official leaderboard
57 gemini-2.0-flash-001 1.72% task version 1.0.0, Epoch Mar 2025 Official leaderboard
58 Qwen Plus Apr 2025 1.72% task version 1.0.0, Epoch Apr 2025 Official leaderboard
59 qwen-max-2025-01-25 1.03% task version 1.0.0, Epoch Apr 2025 Official leaderboard
60 GPT-4.1 nano 1.03% task version 1.0.0, Epoch Apr 2025 Official leaderboard
61 grok-2-1212 0.69% task version 1.0.0, Epoch Mar 2025 Official leaderboard
62 Llama-4-Maverick-17B-128E-Instruct-FP8 0.69% task version 1.0.0, Epoch Apr 2025 Official leaderboard
63 mistral-medium-2505 0.35% task version 1.0.0, Epoch May 2025 Official leaderboard
64 GPT-4o Aug 2024 0.34% task version 1.0.0, Epoch Aug 2024 Official leaderboard
65 Claude 3.5 Haiku 0.34% task version 1.0.0, Epoch Oct 2024 Official leaderboard
66 GPT-4o Nov 2024 0.34% task version 1.0.0, Epoch Nov 2024 Official leaderboard
67 mistral-large-2411 0.34% task version 1.0.0, Epoch Mar 2025 Official leaderboard
68 gemini-1.5-flash-002 0% task version 1.0.0, Epoch Mar 2025 Official leaderboard
69 Llama-4-Scout-17B-16E-Instruct 0% task version 1.0.0, Epoch Apr 2025 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5.5 Pro 52.4% high Apr 2026 Epoch AI run
2 GPT-5.5 51.7% xhigh Apr 2026 Epoch AI run
3 GPT-5.4 Pro 50% xhigh; task version 1.1.5, Epoch Mar 2026 Official leaderboard
4 GPT-5.4 47.6% xhigh; task version 1.1.5, Epoch Mar 2026 Official leaderboard
5 Claude Opus 4.8 47.2% max; task version 1.1.9, Epoch May 2026 Official leaderboard
6 Claude Opus 4.7 43.8% xhigh; task version 1.1.6, Epoch Apr 2026 Official leaderboard
7 GPT-5.2 40.7% xhigh; task version 1.1.2, Epoch Dec 2025 Official leaderboard
8 Claude Opus 4.6 40.7% max; task version 1.1.5, Epoch Feb 2026 Official leaderboard
9 Muse Spark 39% Epoch Apr 2026 Official leaderboard
10 Kimi K2.6 39% task version 1.1.8, Epoch Apr 2026 Official leaderboard
11 Gemini 3.5 Flash 39% high; task version 1.1.8, Epoch May 2026 Official leaderboard
12 Gemini 3 Pro 37.6% task version 1.1.1, Epoch Nov 2025 Official leaderboard
13 Gemini 3.1 Pro 36.9% task version 1.1.5, Epoch Feb 2026 Official leaderboard
14 Gemini 3 Flash 35.6% task version 1.1.2, Epoch Dec 2025 Official leaderboard
15 GLM-5.1 33.5% task version 1.1.8, Epoch Apr 2026 Official leaderboard
16 GPT-5 32.4% high; task version 1.1.0, Epoch Aug 2025 Official leaderboard
17 Claude Sonnet 4.6 32.4% 16K; task version 1.1.5, Epoch Feb 2026 Official leaderboard
18 GPT-5.1 31% high; task version 1.1.0, Epoch Nov 2025 Official leaderboard
19 Gemini 2.5 Deep Think 29% Aug 2025 Epoch AI run
20 GPT-5.4 Mini 28.3% high; task version 1.1.5, Epoch Mar 2026 Official leaderboard
21 Kimi K2.5 27.9% Fireworks Jan 2026 Epoch AI run
22 GPT-5 mini 27.2% high; task version 1.1.0, Epoch Aug 2025 Official leaderboard
23 Qwen 3.6 Plus 26.2% task version 1.1.8, Epoch Mar 2026 Official leaderboard
24 GPT-5.4 Nano 25.9% high; task version 1.1.5, Epoch Mar 2026 Official leaderboard
25 o4-mini 24.8% high; task version 1.1.0, Epoch Apr 2025 Official leaderboard
26 Qwen 3.6 Max (Preview) 23.1% task version 1.1.8, Epoch Apr 2026 Official leaderboard
27 DeepSeek-V3.2 22.1% Thinking; Fireworks Dec 2025 Epoch AI run
28 Kimi K2 Thinking 21.4% task version 1.1.1, Epoch Nov 2025 Official leaderboard
29 Qwen3.5-Plus 21% task version 1.1.8, Epoch Feb 2026 Official leaderboard
30 Qwen 3.5 Plus (hosted 397B-A17B) 21% Feb 2026 Epoch AI run
31 Claude Opus 4.5 20.7% task version 1.1.1, Epoch Nov 2025 Official leaderboard
32 Grok 4 19.7% Jul 2025 Epoch AI run
33 o3 18.7% high; task version 1.1.0, Epoch Apr 2025 Official leaderboard
34 GLM-5 16.4% task version 1.1.5, Epoch Feb 2026 Official leaderboard
35 Claude Sonnet 4.5 15.2% 32K; task version 1.1.0, Epoch Sep 2025 Official leaderboard
36 Gemini 2.5 Pro Jun 2025 14.1% task version 1.1.1, Epoch Jun 2025 Official leaderboard
37 o3-mini 12.4% high; task version 1.1.0, Epoch Jan 2025 Official leaderboard
38 Qwen 3.6 Flash 10.3% task version 1.1.8, Epoch Apr 2026 Official leaderboard
39 o1 9.31% high; task version 1.0.0, Epoch Dec 2024 Official leaderboard
40 Qwen3-235B-A22B-Thinking Jul 2025 8.48% task version 1.1.2, Epoch Jul 2025 Official leaderboard
41 GPT-5 nano 8.28% high Aug 2025 Epoch AI run
42 Claude Opus 4.1 7.24% 27K; task version 1.0.0, Epoch Aug 2025 Official leaderboard
43 Qwen 3.5 Flash (hosted 35B-A3B) 6.21% Feb 2026 Epoch AI run
44 Qwen3.5-Flash 6.21% task version 1.1.8, Epoch Feb 2026 Official leaderboard
45 Claude Haiku 4.5 5.90% 32K; task version 1.0.0, Epoch Oct 2025 Official leaderboard
46 Grok-3 mini 5.86% high; task version 1.0.0, Epoch Apr 2025 Official leaderboard
47 GPT-4.1 5.52% Apr 2025 Epoch AI run
48 Gemini 2.5 Flash Sep 2025 4.84% task version 1.1.2, Epoch Sep 2025 Official leaderboard
49 Gemini 2.5 Flash Jun 2025 4.84% Jun 2025 Epoch AI run
50 GPT-4.1 mini 4.48% task version 1.0.0, Epoch Apr 2025 Official leaderboard
51 Claude Opus 4 4.48% task version 1.0.0, Epoch May 2025 Official leaderboard
52 Claude 3.7 Sonnet 4.14% 16k thinking Feb 2025 Epoch AI run
53 Claude Sonnet 4 4.14% May 2025 Epoch AI run
54 GLM-4.6 3.82% Together Sep 2025 Epoch AI run
55 Grok 3 3.79% task version 1.0.0, Epoch Apr 2025 Official leaderboard
56 GLM-4.7 2.44% Together Dec 2025 Epoch AI run
57 Claude 3.5 Sonnet (October 2024) 2.07% Oct 2024 Epoch AI run
58 Claude 3.5 Sonnet 2.07% task version 1.0.0, Epoch Jun 2024 Official leaderboard
59 o1-mini 1.72% medium; task version 1.0.0, Epoch Sep 2024 Official leaderboard
60 DeepSeek-V3 1.72% task version 1.0.0, Epoch Dec 2024 Official leaderboard
61 Qwen Plus Apr 2025 1.72% task version 1.0.0, Epoch Apr 2025 Official leaderboard
62 Gemini 2.0 Flash Feb 2025 1.72% Feb 2025 Epoch AI run
63 GPT-4.1 nano 1.03% task version 1.0.0, Epoch Apr 2025 Official leaderboard
64 Qwen2.5-Max 1.03% Jan 2025 Epoch AI run
65 Grok-2 Dec 2024 0.69% Dec 2024 Epoch AI run
66 Llama 4 Maverick 0.69% FP8 Apr 2025 Epoch AI run
67 Mistral Medium 3 0.35% May 2025 Epoch AI run
68 GPT-4o Aug 2024 0.34% task version 1.0.0, Epoch Aug 2024 Official leaderboard
69 Claude 3.5 Haiku 0.34% task version 1.0.0, Epoch Oct 2024 Official leaderboard
70 GPT-4o Nov 2024 0.34% task version 1.0.0, Epoch Nov 2024 Official leaderboard
71 Mistral Large 2 Nov 2024 0.34% Nov 2024 Epoch AI run
72 Gemini 1.5 Flash Sep 2024 0% Sep 2024 Epoch AI run
73 Llama 4 Scout 0% Apr 2025 Epoch AI run
#ModelScoreSettingDateSource
1 GPT-5.5 Pro 52.4% high Apr 2026 Epoch AI run
2 GPT-5.5 52.4% Apr 2026 Lab model card
3 GPT-5.4 Pro 50% xhigh; task version 1.1.5, Epoch Mar 2026 Official leaderboard
4 GPT-5.4 47.6% xhigh; task version 1.1.5, Epoch Mar 2026 Official leaderboard
5 Claude Opus 4.8 47.2% max; task version 1.1.9, Epoch May 2026 Official leaderboard
6 Claude Opus 4.7 43.8% xhigh; task version 1.1.6, Epoch Apr 2026 Official leaderboard
7 GPT-5.2 40.7% xhigh; task version 1.1.2, Epoch Dec 2025 Official leaderboard
8 Claude Opus 4.6 40.7% max; task version 1.1.5, Epoch Feb 2026 Official leaderboard
9 Muse Spark 39% Epoch Apr 2026 Official leaderboard
10 Kimi K2.6 39% task version 1.1.8, Epoch Apr 2026 Official leaderboard
11 Gemini 3.5 Flash 39% high; task version 1.1.8, Epoch May 2026 Official leaderboard
12 Gemini 3 Pro 37.6% task version 1.1.1, Epoch Nov 2025 Official leaderboard
13 Gemini 3.1 Pro 36.9% task version 1.1.5, Epoch Feb 2026 Official leaderboard
14 Gemini 3 Flash 35.6% task version 1.1.2, Epoch Dec 2025 Official leaderboard
15 GLM-5.1 33.5% task version 1.1.8, Epoch Apr 2026 Official leaderboard
16 GPT-5 32.4% high; task version 1.1.0, Epoch Aug 2025 Official leaderboard
17 Claude Sonnet 4.6 32.4% 16K; task version 1.1.5, Epoch Feb 2026 Official leaderboard
18 GPT-5.1 31% high; task version 1.1.0, Epoch Nov 2025 Official leaderboard
19 Gemini 2.5 Deep Think 29% Aug 2025 Epoch AI run
20 GPT-5.4 Mini 28.3% high; task version 1.1.5, Epoch Mar 2026 Official leaderboard
21 Kimi K2.5 27.9% Fireworks Jan 2026 Epoch AI run
22 ChatGPT agent 27.4% Jul 2025 Lab model card
23 GPT-5 mini 27.2% high; task version 1.1.0, Epoch Aug 2025 Official leaderboard
24 Qwen 3.6 Plus 26.2% task version 1.1.8, Epoch Mar 2026 Official leaderboard
25 GPT-5.4 Nano 25.9% high; task version 1.1.5, Epoch Mar 2026 Official leaderboard
26 o4-mini 24.8% high; task version 1.1.0, Epoch Apr 2025 Official leaderboard
27 Qwen 3.6 Max (Preview) 23.1% task version 1.1.8, Epoch Apr 2026 Official leaderboard
28 DeepSeek-V3.2 22.1% Thinking; Fireworks Dec 2025 Epoch AI run
29 Kimi K2 Thinking 21.4% task version 1.1.1, Epoch Nov 2025 Official leaderboard
30 Qwen3.5-Plus 21% task version 1.1.8, Epoch Feb 2026 Official leaderboard
31 Qwen 3.5 Plus (hosted 397B-A17B) 21% Feb 2026 Epoch AI run
32 Claude Opus 4.5 20.7% task version 1.1.1, Epoch Nov 2025 Official leaderboard
33 Grok 4 19.7% Jul 2025 Epoch AI run
34 o3 18.7% high; task version 1.1.0, Epoch Apr 2025 Official leaderboard
35 GLM-5 16.4% task version 1.1.5, Epoch Feb 2026 Official leaderboard
36 Claude Sonnet 4.5 15.2% 32K; task version 1.1.0, Epoch Sep 2025 Official leaderboard
37 Gemini 2.5 Pro Jun 2025 14.1% task version 1.1.1, Epoch Jun 2025 Official leaderboard
38 o3-mini 12.4% high; task version 1.1.0, Epoch Jan 2025 Official leaderboard
39 Qwen 3.6 Flash 10.3% task version 1.1.8, Epoch Apr 2026 Official leaderboard
40 GPT-5 nano 9.60% Aug 2025 Lab model card
41 o1 9.31% high; task version 1.0.0, Epoch Dec 2024 Official leaderboard
42 Qwen3-235B-A22B-Thinking Jul 2025 8.48% task version 1.1.2, Epoch Jul 2025 Official leaderboard
43 Claude Opus 4.1 7.24% 27K; task version 1.0.0, Epoch Aug 2025 Official leaderboard
44 MAI-Code-1-Flash 6.30% Jun 2026 Lab model card
45 Qwen 3.5 Flash (hosted 35B-A3B) 6.21% Feb 2026 Epoch AI run
46 Qwen3.5-Flash 6.21% task version 1.1.8, Epoch Feb 2026 Official leaderboard
47 Claude Haiku 4.5 5.90% 32K; task version 1.0.0, Epoch Oct 2025 Official leaderboard
48 Grok-3 mini 5.86% high; task version 1.0.0, Epoch Apr 2025 Official leaderboard
49 GPT-4.1 5.52% Apr 2025 Epoch AI run
50 Gemini 2.5 Flash Sep 2025 4.84% task version 1.1.2, Epoch Sep 2025 Official leaderboard
51 Gemini 2.5 Flash Jun 2025 4.84% Jun 2025 Epoch AI run
52 GPT-4.1 mini 4.48% task version 1.0.0, Epoch Apr 2025 Official leaderboard
53 Claude Opus 4 4.48% task version 1.0.0, Epoch May 2025 Official leaderboard
54 Claude 3.7 Sonnet 4.14% 16k thinking Feb 2025 Epoch AI run
55 Claude Sonnet 4 4.14% May 2025 Epoch AI run
56 GLM-4.6 3.82% Together Sep 2025 Epoch AI run
57 Grok 3 3.79% task version 1.0.0, Epoch Apr 2025 Official leaderboard
58 GLM-4.7 2.44% Together Dec 2025 Epoch AI run
59 Claude 3.5 Sonnet (October 2024) 2.07% Oct 2024 Epoch AI run
60 Claude 3.5 Sonnet 2.07% task version 1.0.0, Epoch Jun 2024 Official leaderboard
61 o1-mini 1.72% medium; task version 1.0.0, Epoch Sep 2024 Official leaderboard
62 DeepSeek-V3 1.72% task version 1.0.0, Epoch Dec 2024 Official leaderboard
63 Qwen Plus Apr 2025 1.72% task version 1.0.0, Epoch Apr 2025 Official leaderboard
64 Gemini 2.0 Flash Feb 2025 1.72% Feb 2025 Epoch AI run
65 GPT-4.1 nano 1.03% task version 1.0.0, Epoch Apr 2025 Official leaderboard
66 Qwen2.5-Max 1.03% Jan 2025 Epoch AI run
67 Grok-2 Dec 2024 0.69% Dec 2024 Epoch AI run
68 Llama 4 Maverick 0.69% FP8 Apr 2025 Epoch AI run
69 Mistral Medium 3 0.35% May 2025 Epoch AI run
70 GPT-4o Aug 2024 0.34% task version 1.0.0, Epoch Aug 2024 Official leaderboard
71 Claude 3.5 Haiku 0.34% task version 1.0.0, Epoch Oct 2024 Official leaderboard
72 GPT-4o Nov 2024 0.34% task version 1.0.0, Epoch Nov 2024 Official leaderboard
73 Mistral Large 2 Nov 2024 0.34% Nov 2024 Epoch AI run
74 Gemini 1.5 Flash Sep 2024 0% Sep 2024 Epoch AI run
75 Llama 4 Scout 0% Apr 2025 Epoch AI run

Epoch AI has run 70 models. Labs report scores for 12 models in their cards. Switch the scores above to rank them together.

More effort adds a median 7 points

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
GPT-5.2 Epoch
low; task v… medium; tas… xhigh; task…
40.7%
Claude Opus 4.6 Epoch
task versio…
40.7%
GPT-5 Epoch
medium; tas… high; task …
32.4%
GPT-5.1 Epoch
none; task … low; task v… medium; tas… high; task …
31%
GPT-5 mini Epoch
medium; tas… high; task …
27.2%
o4-mini Epoch
low; task v… medium; tas… high; task …
24.8%
Claude Opus 4.5 Epoch
task versio…
20.7%
o3 Epoch
low; task v… medium; tas…
18.7%
Claude Sonnet 4.5 Epoch
task versio… 32K; task v…
15.2%
Gemini 2.5 Pro Epoch
task versio… task versio…
14.1%
o3-mini Epoch
medium; tas… high; task …
12.4%
GPT-5 nano Epoch
medium; tas…
8.28%

Source: Official FrontierMath leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5.4 Pro xhigh; task version 1.1.5 Epoch 50%
GPT-5.4 xhigh; task version 1.1.5 Epoch 47.6%
Claude Opus 4.8 max; task version 1.1.9 Epoch 47.2%
Claude Opus 4.7 xhigh; task version 1.1.6 Epoch 43.8%
GPT-5.2 xhigh; task version 1.1.2 Epoch 40.7%
Claude Opus 4.6 max; task version 1.1.5 Epoch 40.7%
GPT-5.2 high; task version 1.1.2 Epoch 40.3%
Claude Opus 4.6 32K; task version 1.1.4 Epoch 40%
Claude Opus 4.6 64K; task version 1.1.4 Epoch 39.7%
Muse Spark Epoch 39%
Kimi K2.6 task version 1.1.8 Epoch 39%
Gemini 3.5 Flash high; task version 1.1.8 Epoch 39%
Claude Opus 4.6 task version 1.1.4 Epoch 38.3%
Gemini 3 Pro task version 1.1.1 Epoch 37.6%
GPT-5.2 medium; task version 1.1.2 Epoch 36.9%
Gemini 3.1 Pro task version 1.1.5 Epoch 36.9%
Gemini 3 Flash task version 1.1.2 Epoch 35.6%
GLM-5.1 task version 1.1.8 Epoch 33.5%
GPT-5 high; task version 1.1.0 Epoch 32.4%
Claude Sonnet 4.6 16K; task version 1.1.5 Epoch 32.4%
GPT-5.1 high; task version 1.1.0 Epoch 31%
GPT-5.4 Mini high; task version 1.1.5 Epoch 28.3%
GPT-5 medium; task version 1.1.0 Epoch 27.2%
GPT-5 mini high; task version 1.1.0 Epoch 27.2%
GPT-5.1 medium; task version 1.1.0 Epoch 26.9%
GPT-5.2 low; task version 1.1.2 Epoch 26.6%
Qwen 3.6 Plus task version 1.1.8 Epoch 26.2%
GPT-5.4 Nano high; task version 1.1.5 Epoch 25.9%
o4-mini high; task version 1.1.0 Epoch 24.8%
Qwen 3.6 Max (Preview) task version 1.1.8 Epoch 23.1%
Kimi K2 Thinking task version 1.1.1 Epoch 21.4%
Qwen3.5-Plus task version 1.1.8 Epoch 21%
Claude Opus 4.5 task version 1.1.1 Epoch 20.7%
Claude Opus 4.5 32K; task version 1.1.1 Epoch 20.7%
GPT-5 mini medium; task version 1.1.0 Epoch 20.3%
Claude Opus 4.5 16K; task version 1.1.1 Epoch 20.3%
o4-mini medium; task version 1.1.0 Epoch 19%
o3 high; task version 1.1.0 Epoch 18.7%
GPT-5.1 low; task version 1.1.1 Epoch 17.3%
o3 medium; task version 1.1.0 Epoch 16.9%
GLM-5 task version 1.1.5 Epoch 16.4%
Claude Sonnet 4.5 32K; task version 1.1.0 Epoch 15.2%
Gemini 2.5 Pro Jun 2025 task version 1.1.1 Epoch 14.1%
Claude Sonnet 4.5 59K; task version 1.1.0 Epoch 13.5%
o3-mini high; task version 1.1.0 Epoch 12.4%
o4-mini low; task version 1.1.0 Epoch 10.7%
Gemini 2.5 Pro Jun 2025 task version 1.0.0 Epoch 10.3%
Qwen 3.6 Flash task version 1.1.8 Epoch 10.3%
o3 low; task version 1.1.0 Epoch 9.72%
o1 high; task version 1.0.0 Epoch 9.31%
Claude Sonnet 4.5 task version 1.1.0 Epoch 9.31%
Qwen3-235B-A22B-Thinking Jul 2025 task version 1.1.2 Epoch 8.48%
GPT-5 nano high; task version 1.0.1 Epoch 8.28%
o3-mini medium; task version 1.0.0 Epoch 8.08%
Claude Opus 4.1 27K; task version 1.0.0 Epoch 7.24%
GPT-5 nano medium; task version 1.0.0 Epoch 7.24%
Qwen3.5-Flash task version 1.1.8 Epoch 6.21%
Claude Haiku 4.5 32K; task version 1.0.0 Epoch 5.90%
Grok-3 mini high; task version 1.0.0 Epoch 5.86%
Claude Opus 4.1 task version 1.0.0 Epoch 5.86%
GPT-4.1 task version 1.0.0 Epoch 5.52%
Gemini 2.5 Flash Sep 2025 task version 1.1.2 Epoch 4.84%
GPT-4.1 mini task version 1.0.0 Epoch 4.48%
Claude Opus 4 task version 1.0.0 Epoch 4.48%
Claude 3.7 Sonnet 16K; task version 1.0.0 Epoch 4.14%
Claude Opus 4 27K; task version 1.0.0 Epoch 4.14%
Claude Sonnet 4 task version 1.0.0 Epoch 4.14%
Claude Haiku 4.5 task version 1.0.0 Epoch 4.14%
GLM-4.6 task version 1.1.1 Epoch 3.82%
Grok 3 task version 1.0.0 Epoch 3.79%
Claude 3.7 Sonnet 32K; task version 1.0.0 Epoch 3.45%
Claude 3.7 Sonnet 64K; task version 1.0.0 Epoch 3.10%
Claude 3.7 Sonnet task version 1.0.0 Epoch 3.10%
Grok-3 mini low; task version 1.0.0 Epoch 2.76%
GLM-4.7 task version 1.1.4 Epoch 2.44%
GPT-5.1 none; task version 1.1.1 Epoch 2.10%
Claude 3.5 Sonnet task version 1.0.0 Epoch 2.07%
o1-mini medium; task version 1.0.0 Epoch 1.72%
DeepSeek-V3 task version 1.0.0 Epoch 1.72%
Qwen Plus Apr 2025 task version 1.0.0 Epoch 1.72%
o1-mini high; task version 1.0.0 Epoch 1.38%
Claude 3.5 Sonnet task version 1.0.0 Epoch 1.03%
GPT-4.1 nano task version 1.0.0 Epoch 1.03%
GPT-4o Aug 2024 task version 1.0.0 Epoch 0.34%
Claude 3.5 Haiku task version 1.0.0 Epoch 0.34%
GPT-4o Nov 2024 task version 1.0.0 Epoch 0.34%

8 models have held number one on the official leaderboard since its release in November 2024 10 models have held number one across all independent runs since its release in November 2024 11 models have held number one with claims included since its release in November 2024

Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard, Epoch AI’s runs and what labs claim in their cards.

Anthropic OpenAI Google DeepMind
Own leaderboard Replaced by Tiers 1-3 v2, Jun 2026
GPT-5.4 Pro 50%
Epoch AI runs Replaced by Tiers 1-3 v2, Jun 2026
GPT-5.5 Pro 52.4%
Lab claims Replaced by Tiers 1-3 v2, Jun 2026
GPT-5.5 52.4%

Sources: Official FrontierMath leaderboard, Epoch AI runs (CC BY 4.0) and labs’ model cards

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Run by Epoch AI Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Highest possible score GPT-5.5 Pro GPT-5.2 GPT-5 ChatGPT agent o3-mini Claude 3.5 Sonnet (October 2024) 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Highest possible score GPT-5.5 Pro GPT-5.2

Sources: Official FrontierMath leaderboard, Epoch AI runs (CC BY 4.0) and labs’ model cards

Labs headline it less often now, last in June 2026

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

OpenAI
Microsoft
DateLaunch postScoreAs the post writes it
Jun 2, 2026 MAI-Code-1-Flash 6.3 Frontier Math (Tier 1-3)
Apr 23, 2026 GPT-5.5 Pro 52.4 FrontierMath Tier 1-3
Apr 23, 2026 GPT-5.5 52.4 FrontierMath Tier 1–3 (GPT-5.5 Pro)
Mar 5, 2026 GPT-5.4 Pro 50 FrontierMath Tier 1-3
Mar 5, 2026 GPT-5.4 47.6 FrontierMath Tier 1-3 (GPT-5.4)
Dec 11, 2025 GPT-5.2 40.3 FrontierMath (Tier 1–3) Advanced mathematics
Nov 13, 2025 GPT-5.1 26.7 FrontierMath (with Python tool)
Aug 7, 2025 GPT-5 nano 9.6 FrontierMath (with python tool only)
Aug 7, 2025 GPT-5 mini 22.1 FrontierMath (with python tool only)
Aug 7, 2025 GPT-5 32.1 FrontierMath, Tier 1-3, GPT-5 pro (python)
Jul 17, 2025 ChatGPT agent 27.4 FrontierMath Tier 1-3, with tools (text)
Jan 31, 2025 o3-mini 9.2 FrontierMath Pass@1

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
Jun 2, 2026 MAI-Code-1-Flash model card 6.3 Frontier Math Tier 1-3, Acc, version not stated (Core reasoning table) Capabilities

The lab's number comes out ahead in 2 of 7 pairs that differ

Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 3 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
GPT-5 mini 22.1% FrontierMath (with python tool only) 27.2% Leaderboard, high; task version 1.1.0, Epoch −5.1 pts
GPT-5.1 26.7% FrontierMath (with Python tool) 31% Leaderboard, high; task version 1.1.0, Epoch −4.3 pts
o3-mini 9.20% FrontierMath Pass@1 12.4% Leaderboard, high; task version 1.1.0, Epoch −3.2 pts
GPT-5 nano 9.60% FrontierMath (with python tool only) 8.28% Leaderboard, high; task version 1.0.1, Epoch +1.3 pts
GPT-5.5 52.4% FrontierMath Tier 1–3 (GPT-5.5 Pro) 51.7% Epoch AI run, xhigh +0.7 pts
GPT-5.2 40.3% FrontierMath (Tier 1–3) Advanced mathematics 40.7% Leaderboard, xhigh; task version 1.1.2, Epoch −0.4 pts
GPT-5 32.1% FrontierMath, Tier 1-3, GPT-5 pro (python) 32.4% Leaderboard, high; task version 1.1.0, Epoch −0.3 pts
GPT-5.4 47.6% FrontierMath Tier 1-3 (GPT-5.4) 47.6% Leaderboard, xhigh; task version 1.1.5, Epoch Same
GPT-5.4 Pro 50% FrontierMath Tier 1-3 50% Leaderboard, xhigh; task version 1.1.5, Epoch Same
GPT-5.5 Pro 52.4% FrontierMath Tier 1-3 52.4% Epoch AI run, high Same

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

Anthropic OpenAI Google DeepMind
FrontierMath Tiers 1-3 v2 Passed 85% in Jun 2026, by GPT-5.5 Pro
Claude Fable 5.2 95.1%
FrontierMath Replaced by Tiers 1-3 v2, Jun 2026
GPT-5.4 Pro 50%
FrontierMath Tiers 1-3 v2 Passed 85% in Jun 2026, by GPT-5.5 Pro
Claude Fable 5.2 95.1%

Source: Official leaderboards

Replaced by FrontierMath Tiers 1-3 v2

More in Math

All 113 in Math
  • AIME 2026 In use Models solve the 30 integer-answer problems from the 2026 AIME I and II math competitions. Scored by exact answer match. GPT-5.5 100%
  • IMO-AnswerBench In use 400 olympiad problems with short verifiable answers, robustified against memorization. Scored by answer match. GLM-5.2 91%
  • HMMT 2026 In use Models solve the 33 final-answer problems from the February 2026 Harvard-MIT Mathematics Tournament. Scored by exact answer match. GPT-5.5 98.5%
  • PolyMath In use Models solve 9,000 parallel math problems in 18 languages across four difficulty levels, from school math to frontier math. Final answers are checked. Qwen3.7-Max 86.5%
  • MathArena (Apex, IMO, USAMO) In use Evaluates models on newly released math competitions after release. Final-answer or human/LLM-judged proof scores. Claude Opus 4.8 81.3%
  • MathArena Apex Shortlist In use Models solve 47 hard final-answer problems that MathArena selected from 2025 competitions, where the best models scored about 50% at selection. Scored by accuracy over repeated attempts. GPT-5.5 98.4%