Benchmark Almanac

HMMT 2025

GPT-5 100% leads since August 2025
GPT-5.2 100% leads since December 2025
GPT-5 100% leads since August 2025
Fading

Models answer the 30 problems from the February 2025 Harvard-MIT Mathematics Tournament. Final answers are checked against the official answers.

Model cards use the MathArena version of the February contest, often averaged over many samples.

Released
February 2025
Kind
Competition problems with answers
Level
Competition
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Strongest model tested
GPT-5.2 Pro #31 on the capability index
Last in a launch post
August 2026, dropped after it saturated
Last model added
August 2026
Scores refreshed
November 12th, 2026
54
Models scored in lab cards
49
Launch posts list it, from 19 labs
25
System cards report it, from 16 labs

3 models share the top score of 100% among models released by August 2026

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

GPT-5.2 leads with 100% across all independent runs

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

GPT-5 leads with 100%, 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 August 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

54 models. Scroll sideways for the rest.

  1. 100.0 GPT-5
  2. 100.0 Qwen3-Max
  3. 100.0 GPT-5.2 Pro
  4. 99.2 DeepSeek-V3.2-Speciale
  5. 98.4 Step 3.5 Flash
  6. 97.9 GLM-5
  7. 97.5 LongCat-Flash-Thinking-2601
  8. 97.3 Seed2.0 Pro
  9. 97.1 GLM-4.7
  10. 96.7 Grok 4 Heavy
  11. 96.7 Qwen 3.6 Plus
  12. 95.4 Kimi K2.5
  13. 94.8 Qwen3.5 397B-A17B
  14. 94.6 Nemotron-Cascade 2 30B-A3B
  15. 93.9 Grok 4
  16. 93.8 Qwen3.6 27B
  17. 93.7 Nemotron 3 Super
  18. 93.3 Grok 4 Fast
  19. 92.5 DeepSeek-V3.2
  20. 92.0 Qwen3.5-27B
  21. 91.4 Qwen3.5-122B-A10B
  22. 90.7 Qwen 3.6 35B-A3B
  23. 89.7 Seed1.8
  24. 89.4 Kimi K2 Thinking
  25. 89.2 Granite 4.2 30B
  26. 89.0 Qwen3.5-35B-A3B
  27. 87.8 GPT-5 mini
  28. 86.7 Ring-1T
  29. 85.8 Tencent HY 2.0 Think
  30. 85.8 Sarvam 105B
  31. 84.4 MiMo-V2-Flash
  32. 84.2 DeepSeek-V3.1
  33. 83.9 Qwen3-235B-A22B-Thinking
  34. 83.7 LongCat-Flash-Thinking
  35. 83.6 DeepSeek-V3.2-Exp
  36. 83.2 Qwen3.5-9B
  37. 79.6 ERNIE 5.0
  38. 79.4 DeepSeek-R1
  39. 78.2 STEP3-VL-10B
  40. 77.4 Qwen3-VL-235B-A22B
  41. 75.6 GPT-5 nano
  42. 74.0 Qwen3.5-4B
  43. 73.9 Qwen3-Next-80B-A3B
  44. 73.3 Solar Open 100B
  45. 73.3 Sarvam 30B
  46. 72.9 EXAONE 4.0 32B
  47. 70.2 Qwen3-Coder-Next
  48. 70.0 Step 3
  49. 67.5 Ministral 3 14B
  50. 64.9 Falcon H1R 7B
  51. 53.6 Phi-4-Reasoning-plus
  52. 43.8 Phi-4-reasoning
  53. 38.8 Kimi K2
  54. 22.9 Qwen3.5-2B
  1. 1 GPT-5 100%
  2. 2 Qwen3-Max 100%
  3. 3 GPT-5.2 Pro 100%
  4. 4 DeepSeek-V3.2-Speciale 99.2%
  5. 5 Step 3.5 Flash 98.4%
  6. 6 GLM-5 97.9%
  7. 7 LongCat-Flash-Thinking-2601 97.5%
  8. 8 Seed2.0 Pro 97.3%
  9. 9 GLM-4.7 97.1%
  10. 10 Grok 4 Heavy 96.7%
  11. 11 Qwen 3.6 Plus 96.7%
  12. 12 Kimi K2.5 95.4%
  13. 13 Qwen3.5 397B-A17B 94.8%
  14. 14 Nemotron-Cascade 2 30B-A3B 94.6%
  15. 15 Grok 4 93.9%
  16. 16 Qwen3.6 27B 93.8%
  17. 17 Nemotron 3 Super 93.7%
  18. 18 Grok 4 Fast 93.3%
  19. 19 DeepSeek-V3.2 92.5%
  20. 20 Qwen3.5-27B 92%
  21. 21 Qwen3.5-122B-A10B 91.4%
  22. 22 Qwen 3.6 35B-A3B 90.7%
  23. 23 Seed1.8 89.7%
  24. 24 Kimi K2 Thinking 89.4%
  25. 25 Granite 4.2 30B 89.2%
  26. 26 Qwen3.5-35B-A3B 89%
  27. 27 GPT-5 mini 87.8%
  28. 28 Ring-1T 86.7%
  29. 29 Tencent HY 2.0 Think 85.8%
  30. 30 Sarvam 105B 85.8%
  31. 31 MiMo-V2-Flash 84.4%
  32. 32 DeepSeek-V3.1 84.2%
  33. 33 Qwen3-235B-A22B-Thinking Jul 2025 83.9%
  34. 34 LongCat-Flash-Thinking 83.7%
  35. 35 DeepSeek-V3.2-Exp 83.6%
  36. 36 Qwen3.5-9B 83.2%
  37. 37 ERNIE 5.0 79.6%
  38. 38 DeepSeek-R1 May 2025 79.4%
  39. 39 STEP3-VL-10B 78.2%
  40. 40 Qwen3-VL-235B-A22B 77.4%
  41. 41 GPT-5 nano 75.6%
  42. 42 Qwen3.5-4B 74%
  43. 43 Qwen3-Next-80B-A3B 73.9%
  44. 44 Solar Open 100B 73.3%
  45. 45 Sarvam 30B 73.3%
  46. 46 EXAONE 4.0 32B 72.9%
  47. 47 Qwen3-Coder-Next 70.2%
  48. 48 Step 3 70%
  49. 49 Ministral 3 14B 67.5%
  50. 50 Falcon H1R 7B 64.9%
  51. 51 Phi-4-Reasoning-plus 53.6%
  52. 52 Phi-4-reasoning 43.8%
  53. 53 Kimi K2 Jul 2025 38.8%
  54. 54 Qwen3.5-2B 22.9%

Source: Labs’ model cards

53 models. Scroll sideways for the rest.

  1. 100.0 GPT-5.2
  2. 98.3 Step 3.5 Flash
  3. 97.5 Gemini 3 Pro
  4. 97.5 DeepSeek-V3.2-Speciale
  5. 97.5 Gemini 3 Flash
  6. 97.5 GLM-5
  7. 95.0 Grok 4
  8. 93.3 GLM-4.6
  9. 93.3 Kimi K2 Thinking
  10. 93.3 GPT-5.1
  11. 93.3 Kimi K2.5
  12. 92.5 DeepSeek-V3.2
  13. 91.7 Grok 4 Fast
  14. 90.0 gpt-oss-120b
  15. 90.0 DeepSeek-V3.2-Exp
  16. 90.0 Grok 4.1 Fast
  17. 89.2 GPT-5 mini
  18. 88.3 GPT-5
  19. 85.8 DeepSeek-V3.1
  20. 84.2 Falcon H1R 7B
  21. 83.3 o4-mini
  22. 82.5 Gemini 2.5 Pro
  23. 80.8 Gemini 2.5 Pro
  24. 77.5 o3
  25. 77.5 GLM-4.5
  26. 76.7 DeepSeek-R1
  27. 76.7 gpt-oss-20b
  28. 76.7 QED-Nano
  29. 74.2 Grok-3 mini
  30. 74.2 GPT-5 nano
  31. 69.2 GLM-4.5-Air
  32. 67.5 o3-mini
  33. 67.5 Claude Sonnet 4.5
  34. 65.0 K2-Think
  35. 64.2 Gemini 2.5 Flash
  36. 62.5 Qwen3-235B-A22B
  37. 60.0 Claude Opus 4
  38. 50.8 Qwen3-30B-A3B
  39. 48.3 o1
  40. 47.5 QwQ-32B
  41. 41.7 DeepSeek-R1
  42. 35.8 Gemini 2.0 Flash Thinking
  43. 33.3 DeepSeek-R1-Distill-Llama-70B
  44. 33.3 DeepSeek-R1-Distill-Qwen-32B
  45. 31.7 DeepSeek-R1-Distill-Qwen-14B
  46. 31.7 Claude 3.7 Sonnet
  47. 29.2 DeepSeek-V3
  48. 18.3 QwQ-32B-Preview
  49. 13.3 Gemini 2.0 Flash
  50. 11.7 DeepSeek-R1-Distill-Qwen-1.5B
  51. 7.50 Gemini 2.0 Pro
  52. 5.83 GPT-4o
  53. 1.67 Claude 3.5 Sonnet
  1. 1 GPT-5.2 xhigh 100%
  2. 2 Step 3.5 Flash 98.3%
  3. 3 Gemini 3 Pro 97.5%
  4. 4 DeepSeek-V3.2-Speciale 97.5%
  5. 5 Gemini 3 Flash high 97.5%
  6. 6 GLM-5 97.5%
  7. 7 Grok 4 95%
  8. 8 GLM-4.6 93.3%
  9. 9 Kimi K2 Thinking 93.3%
  10. 10 GPT-5.1 high 93.3%
  11. 11 Kimi K2.5 high 93.3%
  12. 12 DeepSeek-V3.2 thinking 92.5%
  13. 13 Grok 4 Fast 91.7%
  14. 14 gpt-oss-120b high 90%
  15. 15 DeepSeek-V3.2-Exp thinking 90%
  16. 16 Grok 4.1 Fast reasoning 90%
  17. 17 GPT-5 mini high 89.2%
  18. 18 GPT-5 high 88.3%
  19. 19 DeepSeek-V3.1 thinking 85.8%
  20. 20 Falcon H1R 7B 84.2%
  21. 21 o4-mini high 83.3%
  22. 22 Gemini 2.5 Pro Mar 2025 82.5%
  23. 23 Gemini 2.5 Pro May 2025 80.8%
  24. 24 o3 high 77.5%
  25. 25 GLM-4.5 77.5%
  26. 26 DeepSeek-R1 May 2025 76.7%
  27. 27 gpt-oss-20b high 76.7%
  28. 28 QED-Nano 76.7%
  29. 29 Grok-3 mini high 74.2%
  30. 30 GPT-5 nano high 74.2%
  31. 31 GLM-4.5-Air 69.2%
  32. 32 o3-mini high 67.5%
  33. 33 Claude Sonnet 4.5 thinking 67.5%
  34. 34 K2-Think 65%
  35. 35 Gemini 2.5 Flash Sep 2025, thinking 64.2%
  36. 36 Qwen3-235B-A22B 62.5%
  37. 37 Claude Opus 4 thinking 60%
  38. 38 Qwen3-30B-A3B 50.8%
  39. 39 o1 medium 48.3%
  40. 40 QwQ-32B 47.5%
  41. 41 DeepSeek-R1 41.7%
  42. 42 Gemini 2.0 Flash Thinking Jan 2025 35.8%
  43. 43 DeepSeek-R1-Distill-Llama-70B 33.3%
  44. 44 DeepSeek-R1-Distill-Qwen-32B 33.3%
  45. 45 DeepSeek-R1-Distill-Qwen-14B 31.7%
  46. 46 Claude 3.7 Sonnet thinking 31.7%
  47. 47 DeepSeek-V3 Mar 2025 29.2%
  48. 48 QwQ-32B-Preview 18.3%
  49. 49 Gemini 2.0 Flash Feb 2025 13.3%
  50. 50 DeepSeek-R1-Distill-Qwen-1.5B 11.7%
  51. 51 Gemini 2.0 Pro 7.50%
  52. 52 GPT-4o Mar 2025 5.83%
  53. 53 Claude 3.5 Sonnet 1.67%

Source: MathArena’s leaderboard

92 models. Scroll sideways for the rest.

  1. 100.0 GPT-5
  2. 100.0 Qwen3-Max
  3. 100.0 GPT-5.2
  4. 100.0 GPT-5.2 Pro
  5. 99.2 DeepSeek-V3.2-Speciale
  6. 98.4 Step 3.5 Flash
  7. 97.9 GLM-5
  8. 97.5 Gemini 3 Pro
  9. 97.5 Gemini 3 Flash
  10. 97.5 LongCat-Flash-Thinking-2601
  11. 97.3 Seed2.0 Pro
  12. 97.1 GLM-4.7
  13. 96.7 Grok 4 Heavy
  14. 96.7 Qwen 3.6 Plus
  15. 95.4 Kimi K2.5
  16. 95.0 Grok 4
  17. 94.8 Qwen3.5 397B-A17B
  18. 94.6 Nemotron-Cascade 2 30B-A3B
  19. 93.8 Qwen3.6 27B
  20. 93.7 Nemotron 3 Super
  21. 93.3 GLM-4.6
  22. 93.3 Kimi K2 Thinking
  23. 93.3 GPT-5.1
  24. 93.3 Grok 4 Fast
  25. 92.5 DeepSeek-V3.2
  26. 92.0 Qwen3.5-27B
  27. 91.4 Qwen3.5-122B-A10B
  28. 90.7 Qwen 3.6 35B-A3B
  29. 90.0 gpt-oss-120b
  30. 90.0 DeepSeek-V3.2-Exp
  31. 90.0 Grok 4.1 Fast
  32. 89.7 Seed1.8
  33. 89.2 GPT-5 mini
  34. 89.2 Granite 4.2 30B
  35. 89.0 Qwen3.5-35B-A3B
  36. 86.7 Ring-1T
  37. 85.8 DeepSeek-V3.1
  38. 85.8 Tencent HY 2.0 Think
  39. 85.8 Sarvam 105B
  40. 84.4 MiMo-V2-Flash
  41. 84.2 Falcon H1R 7B
  42. 83.9 Qwen3-235B-A22B-Thinking
  43. 83.7 LongCat-Flash-Thinking
  44. 83.3 o4-mini
  45. 83.2 Qwen3.5-9B
  46. 82.5 Gemini 2.5 Pro
  47. 80.8 Gemini 2.5 Pro
  48. 79.6 ERNIE 5.0
  49. 79.4 DeepSeek-R1
  50. 78.2 STEP3-VL-10B
  51. 77.5 o3
  52. 77.5 GLM-4.5
  53. 77.4 Qwen3-VL-235B-A22B
  54. 76.7 gpt-oss-20b
  55. 76.7 QED-Nano
  56. 75.6 GPT-5 nano
  57. 74.2 Grok-3 mini
  58. 74.0 Qwen3.5-4B
  59. 73.9 Qwen3-Next-80B-A3B
  60. 73.3 Solar Open 100B
  61. 73.3 Sarvam 30B
  62. 72.9 EXAONE 4.0 32B
  63. 70.2 Qwen3-Coder-Next
  64. 70.0 Step 3
  65. 69.2 GLM-4.5-Air
  66. 67.5 o3-mini
  67. 67.5 Claude Sonnet 4.5
  68. 67.5 Ministral 3 14B
  69. 65.0 K2-Think
  70. 64.2 Gemini 2.5 Flash
  71. 62.5 Qwen3-235B-A22B
  72. 60.0 Claude Opus 4
  73. 53.6 Phi-4-Reasoning-plus
  74. 50.8 Qwen3-30B-A3B
  75. 48.3 o1
  76. 47.5 QwQ-32B
  77. 43.8 Phi-4-reasoning
  78. 41.7 DeepSeek-R1
  79. 38.8 Kimi K2
  80. 35.8 Gemini 2.0 Flash Thinking
  81. 33.3 DeepSeek-R1-Distill-Llama-70B
  82. 33.3 DeepSeek-R1-Distill-Qwen-32B
  83. 31.7 DeepSeek-R1-Distill-Qwen-14B
  84. 31.7 Claude 3.7 Sonnet
  85. 29.2 DeepSeek-V3
  86. 22.9 Qwen3.5-2B
  87. 18.3 QwQ-32B-Preview
  88. 13.3 Gemini 2.0 Flash
  89. 11.7 DeepSeek-R1-Distill-Qwen-1.5B
  90. 7.50 Gemini 2.0 Pro
  91. 5.83 GPT-4o
  92. 1.67 Claude 3.5 Sonnet
  1. 1 GPT-5 100%
  2. 2 Qwen3-Max 100%
  3. 3 GPT-5.2 xhigh 100%
  4. 4 GPT-5.2 Pro 100%
  5. 5 DeepSeek-V3.2-Speciale 99.2%
  6. 6 Step 3.5 Flash 98.4%
  7. 7 GLM-5 97.9%
  8. 8 Gemini 3 Pro 97.5%
  9. 9 Gemini 3 Flash high 97.5%
  10. 10 LongCat-Flash-Thinking-2601 97.5%
  11. 11 Seed2.0 Pro 97.3%
  12. 12 GLM-4.7 97.1%
  13. 13 Grok 4 Heavy 96.7%
  14. 14 Qwen 3.6 Plus 96.7%
  15. 15 Kimi K2.5 95.4%
  16. 16 Grok 4 95%
  17. 17 Qwen3.5 397B-A17B 94.8%
  18. 18 Nemotron-Cascade 2 30B-A3B 94.6%
  19. 19 Qwen3.6 27B 93.8%
  20. 20 Nemotron 3 Super 93.7%
  21. 21 GLM-4.6 93.3%
  22. 22 Kimi K2 Thinking 93.3%
  23. 23 GPT-5.1 high 93.3%
  24. 24 Grok 4 Fast 93.3%
  25. 25 DeepSeek-V3.2 thinking 92.5%
  26. 26 Qwen3.5-27B 92%
  27. 27 Qwen3.5-122B-A10B 91.4%
  28. 28 Qwen 3.6 35B-A3B 90.7%
  29. 29 gpt-oss-120b high 90%
  30. 30 DeepSeek-V3.2-Exp thinking 90%
  31. 31 Grok 4.1 Fast reasoning 90%
  32. 32 Seed1.8 89.7%
  33. 33 GPT-5 mini high 89.2%
  34. 34 Granite 4.2 30B 89.2%
  35. 35 Qwen3.5-35B-A3B 89%
  36. 36 Ring-1T 86.7%
  37. 37 DeepSeek-V3.1 thinking 85.8%
  38. 38 Tencent HY 2.0 Think 85.8%
  39. 39 Sarvam 105B 85.8%
  40. 40 MiMo-V2-Flash 84.4%
  41. 41 Falcon H1R 7B 84.2%
  42. 42 Qwen3-235B-A22B-Thinking Jul 2025 83.9%
  43. 43 LongCat-Flash-Thinking 83.7%
  44. 44 o4-mini high 83.3%
  45. 45 Qwen3.5-9B 83.2%
  46. 46 Gemini 2.5 Pro Mar 2025 82.5%
  47. 47 Gemini 2.5 Pro May 2025 80.8%
  48. 48 ERNIE 5.0 79.6%
  49. 49 DeepSeek-R1 May 2025 79.4%
  50. 50 STEP3-VL-10B 78.2%
  51. 51 o3 high 77.5%
  52. 52 GLM-4.5 77.5%
  53. 53 Qwen3-VL-235B-A22B 77.4%
  54. 54 gpt-oss-20b high 76.7%
  55. 55 QED-Nano 76.7%
  56. 56 GPT-5 nano 75.6%
  57. 57 Grok-3 mini high 74.2%
  58. 58 Qwen3.5-4B 74%
  59. 59 Qwen3-Next-80B-A3B 73.9%
  60. 60 Solar Open 100B 73.3%
  61. 61 Sarvam 30B 73.3%
  62. 62 EXAONE 4.0 32B 72.9%
  63. 63 Qwen3-Coder-Next 70.2%
  64. 64 Step 3 70%
  65. 65 GLM-4.5-Air 69.2%
  66. 66 o3-mini high 67.5%
  67. 67 Claude Sonnet 4.5 thinking 67.5%
  68. 68 Ministral 3 14B 67.5%
  69. 69 K2-Think 65%
  70. 70 Gemini 2.5 Flash Sep 2025, thinking 64.2%
  71. 71 Qwen3-235B-A22B 62.5%
  72. 72 Claude Opus 4 thinking 60%
  73. 73 Phi-4-Reasoning-plus 53.6%
  74. 74 Qwen3-30B-A3B 50.8%
  75. 75 o1 medium 48.3%
  76. 76 QwQ-32B 47.5%
  77. 77 Phi-4-reasoning 43.8%
  78. 78 DeepSeek-R1 41.7%
  79. 79 Kimi K2 Jul 2025 38.8%
  80. 80 Gemini 2.0 Flash Thinking Jan 2025 35.8%
  81. 81 DeepSeek-R1-Distill-Llama-70B 33.3%
  82. 82 DeepSeek-R1-Distill-Qwen-32B 33.3%
  83. 83 DeepSeek-R1-Distill-Qwen-14B 31.7%
  84. 84 Claude 3.7 Sonnet thinking 31.7%
  85. 85 DeepSeek-V3 Mar 2025 29.2%
  86. 86 Qwen3.5-2B 22.9%
  87. 87 QwQ-32B-Preview 18.3%
  88. 88 Gemini 2.0 Flash Feb 2025 13.3%
  89. 89 DeepSeek-R1-Distill-Qwen-1.5B 11.7%
  90. 90 Gemini 2.0 Pro 7.50%
  91. 91 GPT-4o Mar 2025 5.83%
  92. 92 Claude 3.5 Sonnet 1.67%

Sources: Labs’ model cards and MathArena’s leaderboard

#ModelScoreSettingDateSource
1 GPT-5 100% Aug 2025 Lab model card
2 Qwen3-Max 100% Sep 2025 Lab model card
3 GPT-5.2 Pro 100% Dec 2025 Lab model card
4 DeepSeek-V3.2-Speciale 99.2% Dec 2025 Lab model card
5 Step 3.5 Flash 98.4% Feb 2026 Lab model card
6 GLM-5 97.9% Feb 2026 Lab model card
7 LongCat-Flash-Thinking-2601 97.5% Jan 2026 Lab model card
8 Seed2.0 Pro 97.3% Feb 2026 Lab model card
9 GLM-4.7 97.1% Dec 2025 Lab model card
10 Grok 4 Heavy 96.7% Jul 2025 Lab model card
11 Qwen 3.6 Plus 96.7% Apr 2026 Lab model card
12 Kimi K2.5 95.4% Jan 2026 Lab model card
13 Qwen3.5 397B-A17B 94.8% Feb 2026 Lab model card
14 Nemotron-Cascade 2 30B-A3B 94.6% Mar 2026 Lab model card
15 Grok 4 93.9% Jul 2025 Lab model card
16 Qwen3.6 27B 93.8% Apr 2026 Lab model card
17 Nemotron 3 Super 93.7% Mar 2026 Lab model card
18 Grok 4 Fast 93.3% Sep 2025 Lab model card
19 DeepSeek-V3.2 92.5% Dec 2025 Lab model card
20 Qwen3.5-27B 92% Feb 2026 Lab model card
21 Qwen3.5-122B-A10B 91.4% Feb 2026 Lab model card
22 Qwen 3.6 35B-A3B 90.7% Apr 2026 Lab model card
23 Seed1.8 89.7% Dec 2025 Lab model card
24 Kimi K2 Thinking 89.4% Nov 2025 Lab model card
25 Granite 4.2 30B 89.2% Aug 2026 Lab model card
26 Qwen3.5-35B-A3B 89% Feb 2026 Lab model card
27 GPT-5 mini 87.8% Aug 2025 Lab model card
28 Ring-1T 86.7% Oct 2025 Lab model card
29 Tencent HY 2.0 Think 85.8% Dec 2025 Lab model card
30 Sarvam 105B 85.8% Mar 2026 Lab model card
31 MiMo-V2-Flash 84.4% Dec 2025 Lab model card
32 DeepSeek-V3.1 84.2% Aug 2025 Lab model card
33 Qwen3-235B-A22B-Thinking Jul 2025 83.9% Jul 2025 Lab model card
34 LongCat-Flash-Thinking 83.7% Sep 2025 Lab model card
35 DeepSeek-V3.2-Exp 83.6% Sep 2025 Lab model card
36 Qwen3.5-9B 83.2% Mar 2026 Lab model card
37 ERNIE 5.0 79.6% Feb 2026 Lab model card
38 DeepSeek-R1 May 2025 79.4% May 2025 Lab model card
39 STEP3-VL-10B 78.2% Jan 2026 Lab model card
40 Qwen3-VL-235B-A22B 77.4% Sep 2025 Lab model card
41 GPT-5 nano 75.6% Aug 2025 Lab model card
42 Qwen3.5-4B 74% Mar 2026 Lab model card
43 Qwen3-Next-80B-A3B 73.9% Sep 2025 Lab model card
44 Solar Open 100B 73.3% Dec 2025 Lab model card
45 Sarvam 30B 73.3% Mar 2026 Lab model card
46 EXAONE 4.0 32B 72.9% Jul 2025 Lab model card
47 Qwen3-Coder-Next 70.2% Feb 2026 Lab model card
48 Step 3 70% Jul 2025 Lab model card
49 Ministral 3 14B 67.5% Jan 2026 Lab model card
50 Falcon H1R 7B 64.9% Jan 2026 Lab model card
51 Phi-4-Reasoning-plus 53.6% Apr 2025 Lab model card
52 Phi-4-reasoning 43.8% Apr 2025 Lab model card
53 Kimi K2 Jul 2025 38.8% Jul 2025 Lab model card
54 Qwen3.5-2B 22.9% Mar 2026 Lab model card
#ModelScoreSettingDateSource
1 GPT-5.2 100% xhigh Dec 2025 Official leaderboard
2 Step 3.5 Flash 98.3% Feb 2026 Official leaderboard
3 Gemini 3 Pro 97.5% Nov 2025 Official leaderboard
4 DeepSeek-V3.2-Speciale 97.5% Dec 2025 Official leaderboard
5 Gemini 3 Flash 97.5% high Dec 2025 Official leaderboard
6 GLM-5 97.5% Feb 2026 Official leaderboard
7 Grok 4 95% Jul 2025 Official leaderboard
8 GLM-4.6 93.3% Sep 2025 Official leaderboard
9 Kimi K2 Thinking 93.3% Nov 2025 Official leaderboard
10 GPT-5.1 93.3% high Nov 2025 Official leaderboard
11 Kimi K2.5 93.3% high Jan 2026 Official leaderboard
12 DeepSeek-V3.2 92.5% thinking Dec 2025 Official leaderboard
13 Grok 4 Fast 91.7% Sep 2025 Official leaderboard
14 gpt-oss-120b 90% high Aug 2025 Official leaderboard
15 DeepSeek-V3.2-Exp 90% thinking Sep 2025 Official leaderboard
16 Grok 4.1 Fast 90% reasoning Nov 2025 Official leaderboard
17 GPT-5 mini 89.2% high Aug 2025 Official leaderboard
18 GPT-5 88.3% high Aug 2025 Official leaderboard
19 DeepSeek-V3.1 85.8% thinking Aug 2025 Official leaderboard
20 Falcon H1R 7B 84.2% Jan 2026 Official leaderboard
21 o4-mini 83.3% high Apr 2025 Official leaderboard
22 Gemini 2.5 Pro Mar 2025 82.5% Mar 2025 Official leaderboard
23 Gemini 2.5 Pro May 2025 80.8% May 2025 Official leaderboard
24 o3 77.5% high Apr 2025 Official leaderboard
25 GLM-4.5 77.5% Aug 2025 Official leaderboard
26 DeepSeek-R1 May 2025 76.7% May 2025 Official leaderboard
27 gpt-oss-20b 76.7% high Aug 2025 Official leaderboard
28 QED-Nano 76.7% Feb 2026 Official leaderboard
29 Grok-3 mini 74.2% high Apr 2025 Official leaderboard
30 GPT-5 nano 74.2% high Aug 2025 Official leaderboard
31 GLM-4.5-Air 69.2% Jul 2025 Official leaderboard
32 o3-mini 67.5% high Jan 2025 Official leaderboard
33 Claude Sonnet 4.5 67.5% thinking Sep 2025 Official leaderboard
34 K2-Think 65% Sep 2025 Official leaderboard
35 Gemini 2.5 Flash Sep 2025 64.2% thinking Sep 2025 Official leaderboard
36 Qwen3-235B-A22B 62.5% Apr 2025 Official leaderboard
37 Claude Opus 4 60% thinking May 2025 Official leaderboard
38 Qwen3-30B-A3B 50.8% Apr 2025 Official leaderboard
39 o1 48.3% medium Dec 2024 Official leaderboard
40 QwQ-32B 47.5% Mar 2025 Official leaderboard
41 DeepSeek-R1 41.7% Jan 2025 Official leaderboard
42 Gemini 2.0 Flash Thinking Jan 2025 35.8% Jan 2025 Official leaderboard
43 DeepSeek-R1-Distill-Llama-70B 33.3% Jan 2025 Official leaderboard
44 DeepSeek-R1-Distill-Qwen-32B 33.3% Jan 2025 Official leaderboard
45 DeepSeek-R1-Distill-Qwen-14B 31.7% Jan 2025 Official leaderboard
46 Claude 3.7 Sonnet 31.7% thinking Feb 2025 Official leaderboard
47 DeepSeek-V3 Mar 2025 29.2% Mar 2025 Official leaderboard
48 QwQ-32B-Preview 18.3% Nov 2024 Official leaderboard
49 Gemini 2.0 Flash Feb 2025 13.3% Feb 2025 Official leaderboard
50 DeepSeek-R1-Distill-Qwen-1.5B 11.7% Jan 2025 Official leaderboard
51 Gemini 2.0 Pro 7.50% Feb 2025 Official leaderboard
52 GPT-4o Mar 2025 5.83% Mar 2025 Official leaderboard
53 Claude 3.5 Sonnet 1.67% Jun 2024 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5 100% Aug 2025 Lab model card
2 Qwen3-Max 100% Sep 2025 Lab model card
3 GPT-5.2 100% xhigh Dec 2025 Official leaderboard
4 GPT-5.2 Pro 100% Dec 2025 Lab model card
5 DeepSeek-V3.2-Speciale 99.2% Dec 2025 Lab model card
6 Step 3.5 Flash 98.4% Feb 2026 Lab model card
7 GLM-5 97.9% Feb 2026 Lab model card
8 Gemini 3 Pro 97.5% Nov 2025 Official leaderboard
9 Gemini 3 Flash 97.5% high Dec 2025 Official leaderboard
10 LongCat-Flash-Thinking-2601 97.5% Jan 2026 Lab model card
11 Seed2.0 Pro 97.3% Feb 2026 Lab model card
12 GLM-4.7 97.1% Dec 2025 Lab model card
13 Grok 4 Heavy 96.7% Jul 2025 Lab model card
14 Qwen 3.6 Plus 96.7% Apr 2026 Lab model card
15 Kimi K2.5 95.4% Jan 2026 Lab model card
16 Grok 4 95% Jul 2025 Official leaderboard
17 Qwen3.5 397B-A17B 94.8% Feb 2026 Lab model card
18 Nemotron-Cascade 2 30B-A3B 94.6% Mar 2026 Lab model card
19 Qwen3.6 27B 93.8% Apr 2026 Lab model card
20 Nemotron 3 Super 93.7% Mar 2026 Lab model card
21 GLM-4.6 93.3% Sep 2025 Official leaderboard
22 Kimi K2 Thinking 93.3% Nov 2025 Official leaderboard
23 GPT-5.1 93.3% high Nov 2025 Official leaderboard
24 Grok 4 Fast 93.3% Sep 2025 Lab model card
25 DeepSeek-V3.2 92.5% thinking Dec 2025 Official leaderboard
26 Qwen3.5-27B 92% Feb 2026 Lab model card
27 Qwen3.5-122B-A10B 91.4% Feb 2026 Lab model card
28 Qwen 3.6 35B-A3B 90.7% Apr 2026 Lab model card
29 gpt-oss-120b 90% high Aug 2025 Official leaderboard
30 DeepSeek-V3.2-Exp 90% thinking Sep 2025 Official leaderboard
31 Grok 4.1 Fast 90% reasoning Nov 2025 Official leaderboard
32 Seed1.8 89.7% Dec 2025 Lab model card
33 GPT-5 mini 89.2% high Aug 2025 Official leaderboard
34 Granite 4.2 30B 89.2% Aug 2026 Lab model card
35 Qwen3.5-35B-A3B 89% Feb 2026 Lab model card
36 Ring-1T 86.7% Oct 2025 Lab model card
37 DeepSeek-V3.1 85.8% thinking Aug 2025 Official leaderboard
38 Tencent HY 2.0 Think 85.8% Dec 2025 Lab model card
39 Sarvam 105B 85.8% Mar 2026 Lab model card
40 MiMo-V2-Flash 84.4% Dec 2025 Lab model card
41 Falcon H1R 7B 84.2% Jan 2026 Official leaderboard
42 Qwen3-235B-A22B-Thinking Jul 2025 83.9% Jul 2025 Lab model card
43 LongCat-Flash-Thinking 83.7% Sep 2025 Lab model card
44 o4-mini 83.3% high Apr 2025 Official leaderboard
45 Qwen3.5-9B 83.2% Mar 2026 Lab model card
46 Gemini 2.5 Pro Mar 2025 82.5% Mar 2025 Official leaderboard
47 Gemini 2.5 Pro May 2025 80.8% May 2025 Official leaderboard
48 ERNIE 5.0 79.6% Feb 2026 Lab model card
49 DeepSeek-R1 May 2025 79.4% May 2025 Lab model card
50 STEP3-VL-10B 78.2% Jan 2026 Lab model card
51 o3 77.5% high Apr 2025 Official leaderboard
52 GLM-4.5 77.5% Aug 2025 Official leaderboard
53 Qwen3-VL-235B-A22B 77.4% Sep 2025 Lab model card
54 gpt-oss-20b 76.7% high Aug 2025 Official leaderboard
55 QED-Nano 76.7% Feb 2026 Official leaderboard
56 GPT-5 nano 75.6% Aug 2025 Lab model card
57 Grok-3 mini 74.2% high Apr 2025 Official leaderboard
58 Qwen3.5-4B 74% Mar 2026 Lab model card
59 Qwen3-Next-80B-A3B 73.9% Sep 2025 Lab model card
60 Solar Open 100B 73.3% Dec 2025 Lab model card
61 Sarvam 30B 73.3% Mar 2026 Lab model card
62 EXAONE 4.0 32B 72.9% Jul 2025 Lab model card
63 Qwen3-Coder-Next 70.2% Feb 2026 Lab model card
64 Step 3 70% Jul 2025 Lab model card
65 GLM-4.5-Air 69.2% Jul 2025 Official leaderboard
66 o3-mini 67.5% high Jan 2025 Official leaderboard
67 Claude Sonnet 4.5 67.5% thinking Sep 2025 Official leaderboard
68 Ministral 3 14B 67.5% Jan 2026 Lab model card
69 K2-Think 65% Sep 2025 Official leaderboard
70 Gemini 2.5 Flash Sep 2025 64.2% thinking Sep 2025 Official leaderboard
71 Qwen3-235B-A22B 62.5% Apr 2025 Official leaderboard
72 Claude Opus 4 60% thinking May 2025 Official leaderboard
73 Phi-4-Reasoning-plus 53.6% Apr 2025 Lab model card
74 Qwen3-30B-A3B 50.8% Apr 2025 Official leaderboard
75 o1 48.3% medium Dec 2024 Official leaderboard
76 QwQ-32B 47.5% Mar 2025 Official leaderboard
77 Phi-4-reasoning 43.8% Apr 2025 Lab model card
78 DeepSeek-R1 41.7% Jan 2025 Official leaderboard
79 Kimi K2 Jul 2025 38.8% Jul 2025 Lab model card
80 Gemini 2.0 Flash Thinking Jan 2025 35.8% Jan 2025 Official leaderboard
81 DeepSeek-R1-Distill-Llama-70B 33.3% Jan 2025 Official leaderboard
82 DeepSeek-R1-Distill-Qwen-32B 33.3% Jan 2025 Official leaderboard
83 DeepSeek-R1-Distill-Qwen-14B 31.7% Jan 2025 Official leaderboard
84 Claude 3.7 Sonnet 31.7% thinking Feb 2025 Official leaderboard
85 DeepSeek-V3 Mar 2025 29.2% Mar 2025 Official leaderboard
86 Qwen3.5-2B 22.9% Mar 2026 Lab model card
87 QwQ-32B-Preview 18.3% Nov 2024 Official leaderboard
88 Gemini 2.0 Flash Feb 2025 13.3% Feb 2025 Official leaderboard
89 DeepSeek-R1-Distill-Qwen-1.5B 11.7% Jan 2025 Official leaderboard
90 Gemini 2.0 Pro 7.50% Feb 2025 Official leaderboard
91 GPT-4o Mar 2025 5.83% Mar 2025 Official leaderboard
92 Claude 3.5 Sonnet 1.67% Jun 2024 Official leaderboard

MathArena’s leaderboard has 53 models. Switch the scores above to rank them together.

The best score, 100%, costs $0.24 a task

Runs on the cost frontier score higher than any run that's cheaper. DeepSeek-V3.2-Speciale comes within 2.5 points of the best for $0.01 a task, 22 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

DeepSeek Google DeepMind xAI OpenAI StepFun Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 $10 Cost per task, log scale GPT-5.2, xhigh Step 3.5 Flash DeepSeek-V3.2, thinking Grok-3 mini, low DeepSeek-R1-Distill-Qwen-14B Gemini 2.0 Flash 0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 $10 Cost per task, log scale GPT-5.2, xhigh DeepSeek-V3.2, thinking

Source: MathArena’s leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5.2 xhigh 100% $0.24 17,003
GPT-5.2 high 98.3% $0.16 11,164
Step 3.5 Flash 98.3% $0.01 47,820
Gemini 3 Pro 97.5% $0.19 15,918
DeepSeek-V3.2-Speciale 97.5% $0.01 25,870
Gemini 3 Flash high 97.5% $0.06 20,536
GLM-5 97.5% $0.09 28,926
Grok 4 95% $0.22 14,669
GLM-4.6 93.3% $0.05 21,480
Kimi K2 Thinking 93.3% $0.07 28,389
GPT-5.1 high 93.3% $0.22 22,001
Kimi K2.5 high 93.3% $0.08 26,862
DeepSeek-V3.2 thinking 92.5% $0.007 17,382
Grok 4 Fast 91.7% $0.006 12,770
gpt-oss-120b high 90% $0.02 27,304
DeepSeek-V3.2-Exp thinking 90% $0.008 18,393
Grok 4.1 Fast reasoning 90% $0.007 13,404
GPT-5 mini high 89.2% $0.03 16,887
GPT-5 high 88.3% $0.17 16,380
DeepSeek-V3.1 thinking 85.8% $0.04 19,230
Falcon H1R 7B 84.2% $0.006 30,233
o4-mini high 83.3% $0.08 17,637
Gemini 2.5 Pro Mar 2025 82.5% $0.13 12,875
Gemini 2.5 Pro May 2025 80.8% 3,623
o3 high 77.5% $0.12 13,786
GLM-4.5 77.5% $0.06 25,438
DeepSeek-R1 May 2025 76.7% $0.06 25,366
gpt-oss-20b high 76.7% $0.01 54,740
QED-Nano 76.7% 43,216
Grok-3 mini high 74.2% $0.01 21,000
GPT-5 nano high 74.2% $0.01 36,743
GLM-4.5-Air 69.2% $0.03 27,720
o3-mini high 67.5% $0.08 17,660
o4-mini medium 67.5% $0.03 7,298
Claude Sonnet 4.5 thinking 67.5% $0.32 21,410
K2-Think 65%
Gemini 2.5 Flash Sep 2025 thinking 64.2% $0.10 27,168
Qwen3-235B-A22B 62.5% $0.009 15,098
Claude Opus 4 thinking 60% $1.23 16,379
o3-mini medium 53.3% $0.03 7,601
Grok-3 mini low 50.8% $0.003 6,636
Qwen3-30B-A3B 50.8% $0.006 18,539
o1 medium 48.3% $0.89 14,835
QwQ-32B 47.5% $0.02 16,152
o4-mini low 47.5% $0.01 2,663
DeepSeek-R1 41.7% $0.03 12,817
Gemini 2.0 Flash Thinking Jan 2025 35.8% 427
DeepSeek-R1-Distill-Llama-70B 33.3% $0.007 11,898
DeepSeek-R1-Distill-Qwen-32B 33.3% $0.005 15,555
DeepSeek-R1-Distill-Qwen-14B 31.7% $0.002 15,559
Claude 3.7 Sonnet thinking 31.7% $0.39 25,902
DeepSeek-V3 Mar 2025 29.2% $0.005 4,216
o3-mini low 28.3% $0.01 2,661
QwQ-32B-Preview 18.3% $0.01 9,531
Gemini 2.0 Flash Feb 2025 13.3% $0.001 3,312
DeepSeek-V3 Mar 2025 13.3% $0.003 2,453
DeepSeek-R1-Distill-Qwen-1.5B 11.7% $0.004 23,111
Gemini 2.0 Pro 7.50% $0.01 2,252
GPT-4o Mar 2025 5.83% $0.008 769
Claude 3.5 Sonnet 1.67% $0.008 528

4 models have held number one in labs’ own numbers since its release in February 2025 6 models have held number one across all independent runs since its release in February 2025 5 models have held number one with claims included since its release in February 2025

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

OpenAI Google DeepMind xAI Microsoft DeepSeek
MathArena leaderboard Passed 85% in Jul 2025, by Grok 4
GPT-5.2 100%
Lab claims Passed 85% in Jul 2025, by Grok 4 Heavy
GPT-5 100%

Sources: MathArena’s leaderboard and labs’ model cards

XAI’s own number for Grok 4 Heavy passed 85% in July 2025

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

On MathArena’s leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) GPT-5 o4-mini o3-mini o1 QwQ-32B-Preview Claude 3.5 Sonnet 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) GPT-5 o4-mini

Sources: MathArena’s leaderboard and labs’ model cards

Labs headline it less often now, last in August 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.

DeepSeek
xAI
Moonshot AI
LG AI Research
Qwen
StepFun
OpenAI
Meituan
Ant Group
Tencent
Xiaomi Corp
ByteDance Seed
Z.ai
Upstage
Technology Innovation Institute
Baidu
Sarvam AI
NVIDIA
IBM
DateLaunch postScoreAs the post writes it
Aug 25, 2026 Granite 4.2 30B 89.2 HMMT Feb25
Apr 22, 2026 Qwen3.6 27B 93.8 HMMT Feb 25
Apr 15, 2026 Qwen 3.6 35B-A3B 90.7 HMMT Feb 25
Apr 2, 2026 Qwen 3.6 Plus 96.7 HMMT Feb 25
Mar 19, 2026 Nemotron-Cascade 2 30B-A3B 94.6 HMMT Feb25
Mar 11, 2026 Nemotron 3 Super 93.7 HMMT Feb25 (no tools)
Mar 6, 2026 Sarvam 30B 73.3 HMMT Feb 2025
Mar 6, 2026 Sarvam 105B 85.8 HMMT (Feb 25)
Mar 2, 2026 Qwen3.5-2B 22.9 HMMT Feb 25
Mar 2, 2026 Qwen3.5-4B 74 HMMT Feb 25
Mar 2, 2026 Qwen3.5-9B 83.2 HMMT Feb 25
Feb 24, 2026 Qwen3.5-35B-A3B 89 HMMT Feb 25
Feb 24, 2026 Qwen3.5-27B 92 HMMT Feb 25
Feb 24, 2026 Qwen3.5-122B-A10B 91.4 HMMT Feb 25
Feb 16, 2026 Qwen3.5 397B-A17B 94.8 HMMT Feb 25
Feb 14, 2026 Seed2.0 Pro 97.3 HMMT Feb 2025
Feb 6, 2026 ERNIE 5.0 79.6 HMMT 2025
Feb 2, 2026 Step 3.5 Flash 98.4 HMMT 2025 (Feb.)
Jan 27, 2026 Kimi K2.5 95.4 HMMT 2025 (Feb)
Jan 23, 2026 Qwen3-Max 98 HMMT Feb 25
Jan 14, 2026 LongCat-Flash-Thinking-2601 97.5 Best of HMMT-25 (Avg@16) w/ tools 93.4 / Heavy Thinking 97.5
Jan 13, 2026 STEP3-VL-10B 78.2 HMMT 2025 (SeRe)
Jan 5, 2026 Falcon H1R 7B 64.9 HMMT-25
Dec 31, 2025 Solar Open 100B 73.3 HMMT 2025 (Feb)
Dec 22, 2025 GLM-4.7 97.1 HMMT Feb. 2025
Dec 18, 2025 Seed1.8 89.7 HMMT25(Feb)
Dec 16, 2025 MiMo-V2-Flash 84.4 HMMT Feb. 2025
Dec 11, 2025 GPT-5.2 Pro 100 HMMT, Feb 2025 (no tools)
Dec 5, 2025 Tencent HY 2.0 Think 85.8 HMMT2025
Dec 1, 2025 DeepSeek-V3.2 92.5 HMMT Feb 2025 (Pass@1)
Nov 6, 2025 Kimi K2 Thinking 89.4 HMMT25, no tools
Oct 10, 2025 Ring-1T 86.7 HMMT 25 (pass@1)
Sep 29, 2025 DeepSeek-V3.2-Exp 83.6 HMMT 2025
Sep 24, 2025 Qwen3-Max 100 HMMT (Thinking Heavy w/ Python)
Sep 23, 2025 Qwen3-VL-235B-A22B 77.4 HMMT25
Sep 21, 2025 LongCat-Flash-Thinking 83.7 HMMT25(Mean@32)
Sep 19, 2025 Grok 4 Fast 93.3 HMMT 2025 (no tools) pass@1
Sep 11, 2025 Qwen3-Next-80B-A3B 73.9 HMMT25
Aug 21, 2025 DeepSeek-V3.1 84.2 HMMT 2025 (Pass@1)
Aug 7, 2025 GPT-5 nano 75.6 HMMT 2025 (no tools)
Aug 7, 2025 GPT-5 mini 87.8 HMMT 2025 (no tools)
Aug 7, 2025 GPT-5 100 HMMT Harvard-MIT mathematics tournament, GPT-5 pro (python)
Jul 31, 2025 Step 3 70 HMMT25
Jul 25, 2025 Qwen3-235B-A22B-Thinking 83.9 HMMT25
Jul 15, 2025 EXAONE 4.0 32B 72.9 HMMT Feb 2025 (reasoning)
Jul 11, 2025 Kimi K2 38.8 HMMT 2025, Avg@32
Jul 9, 2025 Grok 4 93.9 HMMT 2025, with Python (no tools 90.0)
Jul 9, 2025 Grok 4 Heavy 96.7 HMMT 2025 (Competitive Math) [Grok 4 Heavy, Python]
May 28, 2025 DeepSeek-R1 79.4 HMMT 2025 (Pass@1)

25 system cards report it, from 16 labs

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
Mar 21, 2026 Seed1.8 model card 89.7 HMMT25(Feb) (Table 1). Avg@10, 96.7 with increased thinking (Table 7) Capabilities
Mar 19, 2026 Nemotron-Cascade 2 30B-A3B technical report 94.6 HMMT Feb25 (Table 1) Capabilities
Mar 11, 2026 Nemotron 3 Super technical report 93.7 HMMT Feb25 (no tools), 94.73 with tools (Table 5) Capabilities
Feb 28, 2026 Qwen3-Coder-Next technical report 70.2 HMMT25 Feb (Table 9) Capabilities
Feb 17, 2026 GLM-5 technical report 97.9 HMMT Feb. 2025 (Table 7) Capabilities
Feb 14, 2026 Seed2.0 Pro model card 97.3 HMMT Feb 2025 (Table 3) Capabilities
Feb 11, 2026 Step 3.5 Flash technical report 98.4 HMMT 2025 Feb., avg of 64 (Table 5). PaCoRe 100.0, with Python 98.7 (Table 12) Capabilities
Feb 4, 2026 ERNIE 5.0 technical report 79.6 HMMT 2025 (Table 2) Capabilities
Jan 30, 2026 Kimi K2.5 technical report 95.4 HMMT 2025 (Feb), Avg@64 (Table 4) Capabilities
Jan 23, 2026 LongCat-Flash-Thinking-2601 technical report 97.5 HMMT-25 (February) (Avg@16), w/ tools, heavy mode (Table 2). Default mode 93.4 Capabilities
Jan 14, 2026 STEP3-VL-10B technical report 78.2 HMMT25, SeRe (Tables 2 and 3). PaCoRe 92.14 Capabilities
Jan 13, 2026 Ministral 3 14B technical report 67.5 HMMT 2025, reported as pass@16 (Table 5) Capabilities
Jan 11, 2026 Solar Open 100B technical report 73.3 HMMT 2025 (Feb) (Table 6) Capabilities
Jan 6, 2026 MiMo-V2-Flash technical report 84.4 HMMT Feb. 2025 (Table 9) Capabilities
Jan 5, 2026 Falcon H1R 7B technical report 64.9 HMMT25 pass@1 (Table 4) Capabilities
Dec 1, 2025 DeepSeek-V3.2 technical report 92.5 HMMT Feb 2025 (Pass@1), DeepSeek-V3.2 Thinking (Table 2) Capabilities
Dec 1, 2025 DeepSeek-V3.2-Speciale technical report 99.2 HMMT Feb 2025 (Pass@1), DeepSeek-V3.2-Speciale (Table 3) Capabilities
Nov 26, 2025 Qwen3-VL-235B-A22B technical report 77.4 HMMT-25 (Table 6) Capabilities
Oct 21, 2025 Ring-1T technical report 86.7 HMMT25 (Avg@16) (Table 1) Capabilities
Sep 29, 2025 DeepSeek-V3.2-Exp technical report 83.6 HMMT 2025 (Pass@1), DeepSeek-V3.2-Exp (Table 1) Capabilities
Sep 23, 2025 LongCat-Flash-Thinking technical report 83.7 HMMT-25(Mean@32) (Table 2) Capabilities
Jul 21, 2025 Kimi K2 technical report 38.8 HMMT 2025 (Avg@32) (Table 3) Capabilities
Jul 15, 2025 EXAONE 4.0 32B technical report 72.9 HMMT Feb 2025, n=32 (Table 3, Reasoning). Non-reasoning 21.8 (Table 4) Capabilities
Apr 30, 2025 Phi-4-reasoning technical report 43.8 HMMT Feb 2025, average pass@1, MathArena scoring, std 6.2 (Table 1) Capabilities
Apr 30, 2025 Phi-4-Reasoning-plus technical report 53.6 HMMT Feb 2025, average pass@1 over 64 runs, MathArena scoring (Table 1) Capabilities

The lab's number comes out ahead in 7 of 13 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. 2 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
Falcon H1R 7B 64.9% HMMT-25 84.2% Leaderboard −19 pts
GPT-5 100% HMMT Harvard-MIT mathematics tournament, GPT-5 pro (python) 88.3% Leaderboard, high +12 pts
DeepSeek-V3.2-Exp 83.6% HMMT 2025 90% Leaderboard, thinking −6.4 pts
Kimi K2 Thinking 89.4% HMMT25, no tools 93.3% Leaderboard −3.9 pts
DeepSeek-R1 May 2025 79.4% HMMT 2025 (Pass@1) 76.7% Leaderboard +2.7 pts
Kimi K2.5 95.4% HMMT 2025 (Feb) 93.3% Leaderboard, high +2.1 pts
DeepSeek-V3.2-Speciale 99.2% HMMT Feb 2025 (Pass@1), DeepSeek-V3.2-Speciale (Table 3) 97.5% Leaderboard +1.7 pts
DeepSeek-V3.1 84.2% HMMT 2025 (Pass@1) 85.8% Leaderboard, thinking −1.6 pts
Grok 4 Fast 93.3% HMMT 2025 (no tools) pass@1 91.7% Leaderboard +1.6 pts
GPT-5 nano 75.6% HMMT 2025 (no tools) 74.2% Leaderboard, high +1.4 pts
GPT-5 mini 87.8% HMMT 2025 (no tools) 89.2% Leaderboard, high −1.4 pts
Grok 4 93.9% HMMT 2025, with Python (no tools 90.0) 95% Leaderboard −1.1 pts
GLM-5 97.9% HMMT Feb. 2025 (Table 7) 97.5% Leaderboard +0.4 pts
Step 3.5 Flash 98.4% HMMT 2025 (Feb.) 98.3% Leaderboard Same
DeepSeek-V3.2 92.5% HMMT Feb 2025 (Pass@1) 92.5% Leaderboard, thinking 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.

Microsoft DeepSeek xAI OpenAI
HMMT 2026 Passed 85% in Feb 2026, by GPT-5.2
GPT-5.5 98.5%
HMMT 2025 Passed 85% in Jul 2025, by Grok 4 Heavy
GPT-5 100%
HMMT 2026 Passed 85% in Feb 2026, by GPT-5.2
GPT-5.5 98.5%

Sources: Official leaderboards and labs’ model cards

Replaces HMMT February 2024 Replaced by HMMT 2026

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%
  • 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%
  • FrontierMath Tiers 1-3 v2 In use Models solve 295 unpublished math problems from undergraduate to advanced graduate level and submit answers that code checks automatically. Scored by share correct. Claude Fable 5.2 95.1%