Benchmark Almanac

MATH-Vision (MathVision)

Qwen 3.8 Max 97.7% leads since August 2026
GPT-5.4 96.1% leads since March 2026
Qwen 3.8 Max 97.7% leads since August 2026
In use

The official leaderboard hasn’t scored a model released after Seed2.0 Lite (0428) (April 2026). Labs report their own numbers for 14 newer models, such as Qwen 3.8 Max at 97.7%.

3,040 competition math problems with visual context. Answer accuracy.

Made by
CUHK MMLab
Released
February 2024
Kind
Visual math and diagrams
Level
Competition
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Subject
Mathematics
Language
English
In and out
Text, Images → Text
Interaction
Single turn
Items from
Exams and contests
Refresh
One fixed set
Run by
Anyone
In views
Vision
Strongest model tested
Kimi K3 #25 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 20th, 2026

In the authors’ words

Recent advancements in Large Multimodal Models (LMMs) have shown promising results in mathematical reasoning within visual contexts, with models approaching human-level performance on existing benchmarks such as MathVista. However, we observe significant limitations in the diversity of questions and breadth of subjects covered by these benchmarks. To address this issue, we present the MATH-Vision (MATH-V) dataset, a meticulously curated collection of 3,040 high-quality mathematical problems with visual contexts sourced from real math competitions. Spanning 16 distinct mathematical disciplines and graded across 5 levels of difficulty, our dataset provides a comprehensive and diverse set of challenges for evaluating the mathematical reasoning abilities of LMMs. Through extensive experimentation, we unveil a notable performance gap between current LMMs and human performance on MATH-V, underscoring the imperative for further advancements in LMMs. Moreover, our detailed categorization allows for a thorough error analysis of LMMs, offering valuable insights to guide future research and development. The project is available at https://mathvision-cuhk.github.io

Abstract of Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset (arXiv:2402.14804, CC0)
53
Models scored in lab cards
48
Launch posts list it, from 14 labs
32
System cards report it, from 12 labs

Qwen 3.8 Max leads with 97.7%

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

GPT-5.4 leads with 96.1% across all independent runs

Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Dotted bars are reruns by Epoch AI or another board, and each row names its source.

Qwen 3.8 Max leads with 97.7%, 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. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.

  1. 1 Qwen 3.8 Max 97.7%
  2. 2 LongCat-2.5 96.5%
  3. 3 Qwen3.8-Omni-Flash 96.2%
  4. 4 Qwen3.8 Flash 95.7%
  5. 5 Qwen 3.8 27B 94.6%
  6. 6 Seed2.1 Pro 94.5%
  7. 7 Kimi K3 94.3%
  8. 8 Inkling 2 Small 92.8%
  9. 9 Qwen3.7-Plus 90.3%
  10. 10 Seed2.0 Lite (0428) 89.8%
  11. 11 Inkling 2 Mini 89%
  12. 12 Seed2.0 Pro 88.8%
  13. 13 Qwen3.5 397B-A17B 88.6%
  14. 14 Qwen 3.6 Plus 88%
  15. 15 Kimi K2.6 87.4%
  16. 16 Qwen3.5-122B-A10B 86.2%
  17. 17 Qwen3.5-27B 86%
  18. 18 Gemma 4 31B IT 85.6%
  19. 19 Kimi K2.5 84.2%
  20. 20 Qwen3.5-35B-A3B 83.9%
  21. 21 Gemma 4 26B A4B 82.4%
  22. 22 Seed1.8 81.3%
  23. 23 Gemma 4 12B 79.7%
  24. 24 Qwen3.5-9B 78.9%
  25. 25 Nemotron 3.5 Super VL 78.2%
  26. 26 EXAONE 4.5 33B 75.2%
  27. 27 Qwen3-VL-235B-A22B 74.6%
  28. 28 Qwen3.5-4B 74.6%
  29. 29 ERNIE 5.0 74.3%
  30. 30 MiniCPM-V-4.7-35B-A3B 74.2%
  31. 31 Qwen3.5-Omni-Plus 73%
  32. 32 STEP3-VL-10B 70.8%
  33. 33 DiffusionGemma 26B A4B 70.5%
  34. 34 Seed1.6-Thinking 70.3%
  35. 35 Seed1.5-VL 68.7%
  36. 36 GLM-4.5V 65.6%
  37. 37 Step 3 64.8%
  38. 38 InternVL3.5-241B-A28B 63.9%
  39. 39 GLM-4.6V 63.5%
  40. 40 Intern-S1 62.5%
  41. 41 ERNIE-4.5-VL-28B-A3B-Thinking 61.9%
  42. 42 Gemma 4 E4B 59.5%
  43. 43 Kimi-VL-A3B-Thinking-2506 56.9%
  44. 44 Qwen3-Omni-30B-A3B 56.3%
  45. 45 GLM-4.1V-9B-Thinking 54.4%
  46. 46 Nemotron Nano 12B v2 VL 53.6%
  47. 47 Gemma 4 E2B 52.4%
  48. 48 Nemotron 3.5 Bolt E4B 46.3%
  49. 49 Kimi k1.5 38.6%
  50. 50 Qwen2.5-VL-72B 38.1%
  51. 51 Kimi-VL-A3B-Thinking 36.8%
  52. 52 Phi-4-Reasoning-Vision-15B 36.2%
  53. 53 Qwen2.5-Omni-7B 25%

Source: Labs’ model cards

105 models. Scroll sideways for the rest.

  1. 96.1 GPT-5.4
  2. 95.7 Gemini 3.1 Pro
  3. 93.2 Kimi K2.6
  4. 90.6 Gemini 3 Flash
  5. 89.8 Seed2.0 Lite (0428)
  6. 88.8 Seed2.0 Pro
  7. 88.6 Qwen3.5 397B-A17B
  8. 86.6 Gemini 3 Pro
  9. 86.4 Seed2.0 Lite (0215)
  10. 86.2 Qwen3.5-122B-A10B
  11. 86.0 Qwen3.5-27B
  12. 85.6 Gemma 4 31B IT
  13. 85.0 Kimi K2.5
  14. 84.6 Claude Opus 4.6
  15. 83.9 Qwen3.5-35B-A3B
  16. 83.0 GPT-5.2
  17. 82.4 Gemma 4 26B A4B
  18. 81.3 Seed1.8
  19. 77.2 Seed1.6-Thinking
  20. 77.1 Claude Opus 4.5
  21. 74.6 Qwen3-VL-235B-A22B
  22. 73.3 Gemini 2.5 Pro
  23. 72.0 GPT-5
  24. 71.9 GPT-5 mini
  25. 71.1 Claude Sonnet 4.5
  26. 69.6 dots.vlm1
  27. 68.7 Seed1.5-VL
  28. 66.0 Claude Opus 4.1
  29. 66.0 Qwen3-VL-235B-A22B-Instruct
  30. 65.6 GLM-4.5V
  31. 64.8 Step 3
  32. 63.9 InternVL3.5-241B-A28B
  33. 63.5 GLM-4.6V
  34. 60.4 MiMo-VL-7B-RL
  35. 60.3 o1
  36. 59.6 MiMo-VL-7B-RL-2508
  37. 59.6 Qwen3-VL-8B-Thinking
  38. 59.5 Gemma 4 E4B
  39. 58.6 Claude 3.7 Sonnet
  40. 58.0 o4-mini
  41. 57.9 MiMo-VL-7B-SFT
  42. 56.9 Kimi-VL-A3B-Thinking-2506
  43. 55.7 InternVL3.5-30B-A3B
  44. 54.4 GLM-4.1V-9B-Thinking
  45. 54.0 GLM-4.6V-Flash
  46. 54.0 InternVL3.5-38B
  47. 53.9 Ovis2.5-9B
  48. 52.6 Skywork-R1V3-38B
  49. 52.4 Gemma 4 E2B
  50. 52.0 InternVL3.5-8B
  51. 50.3 Qianfan-VL-70B
  52. 49.7 Skywork-R1V2-38B
  53. 48.6 Doubao-1.5-pro
  54. 48.1 Gemini 2.0 Pro
  55. 47.3 GPT-4.5
  56. 46.0 Gemma 3 27B
  57. 46.0 Keye-VL-8B-Preview
  58. 43.1 InternVL3-78B
  59. 41.3 Gemini 2.0 Flash
  60. 40.0 InternVL3.5-4B
  61. 38.6 Kimi k1.5
  62. 38.1 Qwen2.5-VL-72B
  63. 38.0 Claude 3.5 Sonnet
  64. 37.4 Ovis2.5-2B
  65. 37.0 InternVL3-14B
  66. 36.8 Kimi-VL-A3B-Thinking
  67. 35.9 QVQ-72B-Preview
  68. 34.5 InternVL3-38B
  69. 32.8 Qianfan-VL-8B
  70. 32.2 InternVL2_5-78B
  71. 31.9 Ovis2-34B
  72. 31.8 InternVL2.5-38B
  73. 30.4 GPT-4o
  74. 30.3 GPT-4 Turbo
  75. 30.1 Ovis2-16B
  76. 29.0 InternVL3-8B
  77. 27.6 InternVL3-9B
  78. 27.1 Claude 3 Opus
  79. 25.9 Qwen2-VL-72B-Instruct
  80. 25.3 LLaVA-OneVision 72B
  81. 25.1 Qwen2.5-VL-7B
  82. 23.4 MiniCPM-V-2_6
  83. 23.1 InternVL2.5-26B
  84. 22.8 GPT-4V
  85. 21.7 MiniCPM-o-2_6
  86. 21.7 InternVL3-2B
  87. 21.2 Qwen2.5-VL-3B
  88. 19.7 InternVL2.5-8B
  89. 19.2 Gemini 1.5 Pro
  90. 18.8 InternVL3-1B
  91. 18.8 Ovis1.6-Gemma2-9B
  92. 17.9 Aquila-VL-2B
  93. 17.7 Gemini 1.0 Pro
  94. 17.0 InternVL-Chat-V1-2-Plus
  95. 16.3 Qwen2-VL-7B-Instruct
  96. 15.6 Qwen-VL-Max
  97. 14.7 InternVL2.5-2B
  98. 14.5 InternLM-XComposer2-VL-7B
  99. 14.4 InternVL2.5-1B
  100. 12.4 Qwen2-VL-2B-Instruct
  101. 11.9 ShareGPT4V-13B
  102. 11.1 LLaVA-v1.5-13B
  103. 10.7 Qwen-VL-Plus
  104. 10.5 ShareGPT4V-7B
  105. 8.52 llava-v1.5-7b
  1. 1 GPT-5.4 xhigh, with python 96.1%
  2. 2 Gemini 3.1 Pro thinking high, with python 95.7%
  3. 3 Kimi K2.6 with python 93.2%
  4. 4 Gemini 3 Flash 90.6%
  5. 5 Seed2.0 Lite (0428) 89.8%
  6. 6 Seed2.0 Pro 88.8%
  7. 7 Qwen3.5 397B-A17B 88.6%
  8. 8 Gemini 3 Pro 86.6%
  9. 9 Seed2.0 Lite (0215) 86.4%
  10. 10 Qwen3.5-122B-A10B 86.2%
  11. 11 Qwen3.5-27B 86%
  12. 12 Gemma 4 31B IT 85.6%
  13. 13 Kimi K2.5 thinking, with python 85%
  14. 14 Claude Opus 4.6 max, with python 84.6%
  15. 15 Qwen3.5-35B-A3B 83.9%
  16. 16 GPT-5.2 83%
  17. 17 Gemma 4 26B A4B 82.4%
  18. 18 Seed1.8 81.3%
  19. 19 Seed1.6-Thinking 77.2%
  20. 20 Claude Opus 4.5 77.1%
  21. 21 Qwen3-VL-235B-A22B thinking 74.6%
  22. 22 Gemini 2.5 Pro Mar 2025 73.3%
  23. 23 GPT-5 72%
  24. 24 GPT-5 mini 71.9%
  25. 25 Claude Sonnet 4.5 71.1%
  26. 26 dots.vlm1 69.6%
  27. 27 Seed1.5-VL 68.7%
  28. 28 Claude Opus 4.1 thinking 66%
  29. 29 Qwen3-VL-235B-A22B-Instruct 66%
  30. 30 GLM-4.5V 65.6%
  31. 31 Step 3 64.8%
  32. 32 InternVL3.5-241B-A28B 63.9%
  33. 33 GLM-4.6V 63.5%
  34. 34 MiMo-VL-7B-RL 60.4%
  35. 35 o1 60.3%
  36. 36 MiMo-VL-7B-RL-2508 59.6%
  37. 37 Qwen3-VL-8B-Thinking 59.6%
  38. 38 Gemma 4 E4B 59.5%
  39. 39 Claude 3.7 Sonnet 58.6%
  40. 40 o4-mini 58%
  41. 41 MiMo-VL-7B-SFT 57.9%
  42. 42 Kimi-VL-A3B-Thinking-2506 56.9%
  43. 43 InternVL3.5-30B-A3B 55.7%
  44. 44 GLM-4.1V-9B-Thinking 54.4%
  45. 45 GLM-4.6V-Flash 54%
  46. 46 InternVL3.5-38B 54%
  47. 47 Ovis2.5-9B 53.9%
  48. 48 Skywork-R1V3-38B 52.6%
  49. 49 Gemma 4 E2B 52.4%
  50. 50 InternVL3.5-8B 52%
  51. 51 Qianfan-VL-70B 50.3%
  52. 52 Skywork-R1V2-38B 49.7%
  53. 53 Doubao-1.5-pro 48.6%
  54. 54 Gemini 2.0 Pro 48.1%
  55. 55 GPT-4.5 47.3%
  56. 56 Gemma 3 27B non-thinking 46%
  57. 57 Keye-VL-8B-Preview 46%
  58. 58 InternVL3-78B 43.1%
  59. 59 Gemini 2.0 Flash Feb 2025 41.3%
  60. 60 InternVL3.5-4B 40%
  61. 61 Kimi k1.5 38.6%
  62. 62 Qwen2.5-VL-72B 38.1%
  63. 63 Claude 3.5 Sonnet 38%
  64. 64 Ovis2.5-2B 37.4%
  65. 65 InternVL3-14B 37%
  66. 66 Kimi-VL-A3B-Thinking 36.8%
  67. 67 QVQ-72B-Preview 35.9%
  68. 68 InternVL3-38B 34.5%
  69. 69 Qianfan-VL-8B 32.8%
  70. 70 InternVL2_5-78B 32.2%
  71. 71 Ovis2-34B 31.9%
  72. 72 InternVL2.5-38B 31.8%
  73. 73 GPT-4o May 2024 30.4%
  74. 74 GPT-4 Turbo Apr 2024 30.3%
  75. 75 Ovis2-16B 30.1%
  76. 76 InternVL3-8B 29%
  77. 77 InternVL3-9B 27.6%
  78. 78 Claude 3 Opus 27.1%
  79. 79 Qwen2-VL-72B-Instruct 25.9%
  80. 80 LLaVA-OneVision 72B 25.3%
  81. 81 Qwen2.5-VL-7B 25.1%
  82. 82 MiniCPM-V-2_6 23.4%
  83. 83 InternVL2.5-26B 23.1%
  84. 84 GPT-4V 22.8%
  85. 85 MiniCPM-o-2_6 21.7%
  86. 86 InternVL3-2B 21.7%
  87. 87 Qwen2.5-VL-3B 21.2%
  88. 88 InternVL2.5-8B 19.7%
  89. 89 Gemini 1.5 Pro May 2024 19.2%
  90. 90 InternVL3-1B 18.8%
  91. 91 Ovis1.6-Gemma2-9B 18.8%
  92. 92 Aquila-VL-2B 17.9%
  93. 93 Gemini 1.0 Pro 17.7%
  94. 94 InternVL-Chat-V1-2-Plus 17%
  95. 95 Qwen2-VL-7B-Instruct 16.3%
  96. 96 Qwen-VL-Max 15.6%
  97. 97 InternVL2.5-2B 14.7%
  98. 98 InternLM-XComposer2-VL-7B 14.5%
  99. 99 InternVL2.5-1B 14.4%
  100. 100 Qwen2-VL-2B-Instruct 12.4%
  101. 101 ShareGPT4V-13B 11.9%
  102. 102 LLaVA-v1.5-13B 11.1%
  103. 103 Qwen-VL-Plus 10.7%
  104. 104 ShareGPT4V-7B 10.5%
  105. 105 llava-v1.5-7b 8.52%

Source: Official MATH-Vision (MathVision) leaderboard

133 models. Scroll sideways for the rest.

  1. 97.7 Qwen 3.8 Max
  2. 96.5 LongCat-2.5
  3. 96.2 Qwen3.8-Omni-Flash
  4. 96.1 GPT-5.4
  5. 95.7 Gemini 3.1 Pro
  6. 95.7 Qwen3.8 Flash
  7. 94.6 Qwen 3.8 27B
  8. 94.5 Seed2.1 Pro
  9. 94.3 Kimi K3
  10. 93.2 Kimi K2.6
  11. 92.8 Inkling 2 Small
  12. 90.6 Gemini 3 Flash
  13. 90.3 Qwen3.7-Plus
  14. 89.8 Seed2.0 Lite (0428)
  15. 89.0 Inkling 2 Mini
  16. 88.8 Seed2.0 Pro
  17. 88.6 Qwen3.5 397B-A17B
  18. 88.0 Qwen 3.6 Plus
  19. 86.6 Gemini 3 Pro
  20. 86.4 Seed2.0 Lite (0215)
  21. 86.2 Qwen3.5-122B-A10B
  22. 86.0 Qwen3.5-27B
  23. 85.6 Gemma 4 31B IT
  24. 85.0 Kimi K2.5
  25. 84.6 Claude Opus 4.6
  26. 83.9 Qwen3.5-35B-A3B
  27. 83.0 GPT-5.2
  28. 82.4 Gemma 4 26B A4B
  29. 81.3 Seed1.8
  30. 79.7 Gemma 4 12B
  31. 78.9 Qwen3.5-9B
  32. 78.2 Nemotron 3.5 Super VL
  33. 77.2 Seed1.6-Thinking
  34. 77.1 Claude Opus 4.5
  35. 75.2 EXAONE 4.5 33B
  36. 74.6 Qwen3-VL-235B-A22B
  37. 74.6 Qwen3.5-4B
  38. 74.3 ERNIE 5.0
  39. 74.2 MiniCPM-V-4.7-35B-A3B
  40. 73.3 Gemini 2.5 Pro
  41. 73.0 Qwen3.5-Omni-Plus
  42. 72.0 GPT-5
  43. 71.9 GPT-5 mini
  44. 71.1 Claude Sonnet 4.5
  45. 70.8 STEP3-VL-10B
  46. 70.5 DiffusionGemma 26B A4B
  47. 69.6 dots.vlm1
  48. 68.7 Seed1.5-VL
  49. 66.0 Claude Opus 4.1
  50. 66.0 Qwen3-VL-235B-A22B-Instruct
  51. 65.6 GLM-4.5V
  52. 64.8 Step 3
  53. 63.9 InternVL3.5-241B-A28B
  54. 63.5 GLM-4.6V
  55. 62.5 Intern-S1
  56. 61.9 ERNIE-4.5-VL-28B-A3B-Thinking
  57. 60.4 MiMo-VL-7B-RL
  58. 60.3 o1
  59. 59.6 MiMo-VL-7B-RL-2508
  60. 59.6 Qwen3-VL-8B-Thinking
  61. 59.5 Gemma 4 E4B
  62. 58.6 Claude 3.7 Sonnet
  63. 58.0 o4-mini
  64. 57.9 MiMo-VL-7B-SFT
  65. 56.9 Kimi-VL-A3B-Thinking-2506
  66. 56.3 Qwen3-Omni-30B-A3B
  67. 55.7 InternVL3.5-30B-A3B
  68. 54.4 GLM-4.1V-9B-Thinking
  69. 54.0 GLM-4.6V-Flash
  70. 54.0 InternVL3.5-38B
  71. 53.9 Ovis2.5-9B
  72. 53.6 Nemotron Nano 12B v2 VL
  73. 52.6 Skywork-R1V3-38B
  74. 52.4 Gemma 4 E2B
  75. 52.0 InternVL3.5-8B
  76. 50.3 Qianfan-VL-70B
  77. 49.7 Skywork-R1V2-38B
  78. 48.6 Doubao-1.5-pro
  79. 48.1 Gemini 2.0 Pro
  80. 47.3 GPT-4.5
  81. 46.3 Nemotron 3.5 Bolt E4B
  82. 46.0 Gemma 3 27B
  83. 46.0 Keye-VL-8B-Preview
  84. 43.1 InternVL3-78B
  85. 41.3 Gemini 2.0 Flash
  86. 40.0 InternVL3.5-4B
  87. 38.6 Kimi k1.5
  88. 38.1 Qwen2.5-VL-72B
  89. 38.0 Claude 3.5 Sonnet
  90. 37.4 Ovis2.5-2B
  91. 37.0 InternVL3-14B
  92. 36.8 Kimi-VL-A3B-Thinking
  93. 36.2 Phi-4-Reasoning-Vision-15B
  94. 35.9 QVQ-72B-Preview
  95. 34.5 InternVL3-38B
  96. 32.8 Qianfan-VL-8B
  97. 32.2 InternVL2_5-78B
  98. 31.9 Ovis2-34B
  99. 31.8 InternVL2.5-38B
  100. 30.4 GPT-4o
  101. 30.3 GPT-4 Turbo
  102. 30.1 Ovis2-16B
  103. 29.0 InternVL3-8B
  104. 27.6 InternVL3-9B
  105. 27.1 Claude 3 Opus
  106. 25.9 Qwen2-VL-72B-Instruct
  107. 25.3 LLaVA-OneVision 72B
  108. 25.1 Qwen2.5-VL-7B
  109. 25.0 Qwen2.5-Omni-7B
  110. 23.4 MiniCPM-V-2_6
  111. 23.1 InternVL2.5-26B
  112. 22.8 GPT-4V
  113. 21.7 MiniCPM-o-2_6
  114. 21.7 InternVL3-2B
  115. 21.2 Qwen2.5-VL-3B
  116. 19.7 InternVL2.5-8B
  117. 19.2 Gemini 1.5 Pro
  118. 18.8 InternVL3-1B
  119. 18.8 Ovis1.6-Gemma2-9B
  120. 17.9 Aquila-VL-2B
  121. 17.7 Gemini 1.0 Pro
  122. 17.0 InternVL-Chat-V1-2-Plus
  123. 16.3 Qwen2-VL-7B-Instruct
  124. 15.6 Qwen-VL-Max
  125. 14.7 InternVL2.5-2B
  126. 14.5 InternLM-XComposer2-VL-7B
  127. 14.4 InternVL2.5-1B
  128. 12.4 Qwen2-VL-2B-Instruct
  129. 11.9 ShareGPT4V-13B
  130. 11.1 LLaVA-v1.5-13B
  131. 10.7 Qwen-VL-Plus
  132. 10.5 ShareGPT4V-7B
  133. 8.52 llava-v1.5-7b
  1. 1 Qwen 3.8 Max 97.7%
  2. 2 LongCat-2.5 96.5%
  3. 3 Qwen3.8-Omni-Flash 96.2%
  4. 4 GPT-5.4 xhigh, with python 96.1%
  5. 5 Gemini 3.1 Pro thinking high, with python 95.7%
  6. 6 Qwen3.8 Flash 95.7%
  7. 7 Qwen 3.8 27B 94.6%
  8. 8 Seed2.1 Pro 94.5%
  9. 9 Kimi K3 94.3%
  10. 10 Kimi K2.6 with python 93.2%
  11. 11 Inkling 2 Small 92.8%
  12. 12 Gemini 3 Flash 90.6%
  13. 13 Qwen3.7-Plus 90.3%
  14. 14 Seed2.0 Lite (0428) 89.8%
  15. 15 Inkling 2 Mini 89%
  16. 16 Seed2.0 Pro 88.8%
  17. 17 Qwen3.5 397B-A17B 88.6%
  18. 18 Qwen 3.6 Plus 88%
  19. 19 Gemini 3 Pro 86.6%
  20. 20 Seed2.0 Lite (0215) 86.4%
  21. 21 Qwen3.5-122B-A10B 86.2%
  22. 22 Qwen3.5-27B 86%
  23. 23 Gemma 4 31B IT 85.6%
  24. 24 Kimi K2.5 thinking, with python 85%
  25. 25 Claude Opus 4.6 max, with python 84.6%
  26. 26 Qwen3.5-35B-A3B 83.9%
  27. 27 GPT-5.2 83%
  28. 28 Gemma 4 26B A4B 82.4%
  29. 29 Seed1.8 81.3%
  30. 30 Gemma 4 12B 79.7%
  31. 31 Qwen3.5-9B 78.9%
  32. 32 Nemotron 3.5 Super VL 78.2%
  33. 33 Seed1.6-Thinking 77.2%
  34. 34 Claude Opus 4.5 77.1%
  35. 35 EXAONE 4.5 33B 75.2%
  36. 36 Qwen3-VL-235B-A22B thinking 74.6%
  37. 37 Qwen3.5-4B 74.6%
  38. 38 ERNIE 5.0 74.3%
  39. 39 MiniCPM-V-4.7-35B-A3B 74.2%
  40. 40 Gemini 2.5 Pro Mar 2025 73.3%
  41. 41 Qwen3.5-Omni-Plus 73%
  42. 42 GPT-5 72%
  43. 43 GPT-5 mini 71.9%
  44. 44 Claude Sonnet 4.5 71.1%
  45. 45 STEP3-VL-10B 70.8%
  46. 46 DiffusionGemma 26B A4B 70.5%
  47. 47 dots.vlm1 69.6%
  48. 48 Seed1.5-VL 68.7%
  49. 49 Claude Opus 4.1 thinking 66%
  50. 50 Qwen3-VL-235B-A22B-Instruct 66%
  51. 51 GLM-4.5V 65.6%
  52. 52 Step 3 64.8%
  53. 53 InternVL3.5-241B-A28B 63.9%
  54. 54 GLM-4.6V 63.5%
  55. 55 Intern-S1 62.5%
  56. 56 ERNIE-4.5-VL-28B-A3B-Thinking 61.9%
  57. 57 MiMo-VL-7B-RL 60.4%
  58. 58 o1 60.3%
  59. 59 MiMo-VL-7B-RL-2508 59.6%
  60. 60 Qwen3-VL-8B-Thinking 59.6%
  61. 61 Gemma 4 E4B 59.5%
  62. 62 Claude 3.7 Sonnet 58.6%
  63. 63 o4-mini 58%
  64. 64 MiMo-VL-7B-SFT 57.9%
  65. 65 Kimi-VL-A3B-Thinking-2506 56.9%
  66. 66 Qwen3-Omni-30B-A3B 56.3%
  67. 67 InternVL3.5-30B-A3B 55.7%
  68. 68 GLM-4.1V-9B-Thinking 54.4%
  69. 69 GLM-4.6V-Flash 54%
  70. 70 InternVL3.5-38B 54%
  71. 71 Ovis2.5-9B 53.9%
  72. 72 Nemotron Nano 12B v2 VL 53.6%
  73. 73 Skywork-R1V3-38B 52.6%
  74. 74 Gemma 4 E2B 52.4%
  75. 75 InternVL3.5-8B 52%
  76. 76 Qianfan-VL-70B 50.3%
  77. 77 Skywork-R1V2-38B 49.7%
  78. 78 Doubao-1.5-pro 48.6%
  79. 79 Gemini 2.0 Pro 48.1%
  80. 80 GPT-4.5 47.3%
  81. 81 Nemotron 3.5 Bolt E4B 46.3%
  82. 82 Gemma 3 27B non-thinking 46%
  83. 83 Keye-VL-8B-Preview 46%
  84. 84 InternVL3-78B 43.1%
  85. 85 Gemini 2.0 Flash Feb 2025 41.3%
  86. 86 InternVL3.5-4B 40%
  87. 87 Kimi k1.5 38.6%
  88. 88 Qwen2.5-VL-72B 38.1%
  89. 89 Claude 3.5 Sonnet 38%
  90. 90 Ovis2.5-2B 37.4%
  91. 91 InternVL3-14B 37%
  92. 92 Kimi-VL-A3B-Thinking 36.8%
  93. 93 Phi-4-Reasoning-Vision-15B 36.2%
  94. 94 QVQ-72B-Preview 35.9%
  95. 95 InternVL3-38B 34.5%
  96. 96 Qianfan-VL-8B 32.8%
  97. 97 InternVL2_5-78B 32.2%
  98. 98 Ovis2-34B 31.9%
  99. 99 InternVL2.5-38B 31.8%
  100. 100 GPT-4o May 2024 30.4%
  101. 101 GPT-4 Turbo Apr 2024 30.3%
  102. 102 Ovis2-16B 30.1%
  103. 103 InternVL3-8B 29%
  104. 104 InternVL3-9B 27.6%
  105. 105 Claude 3 Opus 27.1%
  106. 106 Qwen2-VL-72B-Instruct 25.9%
  107. 107 LLaVA-OneVision 72B 25.3%
  108. 108 Qwen2.5-VL-7B 25.1%
  109. 109 Qwen2.5-Omni-7B 25%
  110. 110 MiniCPM-V-2_6 23.4%
  111. 111 InternVL2.5-26B 23.1%
  112. 112 GPT-4V 22.8%
  113. 113 MiniCPM-o-2_6 21.7%
  114. 114 InternVL3-2B 21.7%
  115. 115 Qwen2.5-VL-3B 21.2%
  116. 116 InternVL2.5-8B 19.7%
  117. 117 Gemini 1.5 Pro May 2024 19.2%
  118. 118 InternVL3-1B 18.8%
  119. 119 Ovis1.6-Gemma2-9B 18.8%
  120. 120 Aquila-VL-2B 17.9%
  121. 121 Gemini 1.0 Pro 17.7%
  122. 122 InternVL-Chat-V1-2-Plus 17%
  123. 123 Qwen2-VL-7B-Instruct 16.3%
  124. 124 Qwen-VL-Max 15.6%
  125. 125 InternVL2.5-2B 14.7%
  126. 126 InternLM-XComposer2-VL-7B 14.5%
  127. 127 InternVL2.5-1B 14.4%
  128. 128 Qwen2-VL-2B-Instruct 12.4%
  129. 129 ShareGPT4V-13B 11.9%
  130. 130 LLaVA-v1.5-13B 11.1%
  131. 131 Qwen-VL-Plus 10.7%
  132. 132 ShareGPT4V-7B 10.5%
  133. 133 llava-v1.5-7b 8.52%

Sources: Labs’ model cards and official MATH-Vision (MathVision) leaderboard

#ModelScoreSettingDateSource
1 Qwen 3.8 Max 97.7% Aug 2026 Lab model card
2 LongCat-2.5 96.5% Nov 2026 Lab model card
3 Qwen3.8-Omni-Flash 96.2% Sep 2026 Lab model card
4 Qwen3.8 Flash 95.7% Aug 2026 Lab model card
5 Qwen 3.8 27B 94.6% Aug 2026 Lab model card
6 Seed2.1 Pro 94.5% Jun 2026 Lab model card
7 Kimi K3 94.3% Jul 2026 Lab model card
8 Inkling 2 Small 92.8% Nov 2026 Lab model card
9 Qwen3.7-Plus 90.3% Jun 2026 Lab model card
10 Seed2.0 Lite (0428) 89.8% Apr 2026 Lab model card
11 Inkling 2 Mini 89% Nov 2026 Lab model card
12 Seed2.0 Pro 88.8% Feb 2026 Lab model card
13 Qwen3.5 397B-A17B 88.6% Feb 2026 Lab model card
14 Qwen 3.6 Plus 88% Apr 2026 Lab model card
15 Kimi K2.6 87.4% Apr 2026 Lab model card
16 Qwen3.5-122B-A10B 86.2% Feb 2026 Lab model card
17 Qwen3.5-27B 86% Feb 2026 Lab model card
18 Gemma 4 31B IT 85.6% Apr 2026 Lab model card
19 Kimi K2.5 84.2% Jan 2026 Lab model card
20 Qwen3.5-35B-A3B 83.9% Feb 2026 Lab model card
21 Gemma 4 26B A4B 82.4% Apr 2026 Lab model card
22 Seed1.8 81.3% Dec 2025 Lab model card
23 Gemma 4 12B 79.7% Jun 2026 Lab model card
24 Qwen3.5-9B 78.9% Mar 2026 Lab model card
25 Nemotron 3.5 Super VL 78.2% Oct 2026 Lab model card
26 EXAONE 4.5 33B 75.2% Apr 2026 Lab model card
27 Qwen3-VL-235B-A22B 74.6% Sep 2025 Lab model card
28 Qwen3.5-4B 74.6% Mar 2026 Lab model card
29 ERNIE 5.0 74.3% Feb 2026 Lab model card
30 MiniCPM-V-4.7-35B-A3B 74.2% Oct 2026 Lab model card
31 Qwen3.5-Omni-Plus 73% Mar 2026 Lab model card
32 STEP3-VL-10B 70.8% Jan 2026 Lab model card
33 DiffusionGemma 26B A4B 70.5% Jun 2026 Lab model card
34 Seed1.6-Thinking 70.3% Jun 2025 Lab model card
35 Seed1.5-VL 68.7% May 2025 Lab model card
36 GLM-4.5V 65.6% Aug 2025 Lab model card
37 Step 3 64.8% Jul 2025 Lab model card
38 InternVL3.5-241B-A28B 63.9% Aug 2025 Lab model card
39 GLM-4.6V 63.5% Jan 2026 Lab model card
40 Intern-S1 62.5% Jul 2025 Lab model card
41 ERNIE-4.5-VL-28B-A3B-Thinking 61.9% Nov 2025 Lab model card
42 Gemma 4 E4B 59.5% Apr 2026 Lab model card
43 Kimi-VL-A3B-Thinking-2506 56.9% Jun 2025 Lab model card
44 Qwen3-Omni-30B-A3B 56.3% Sep 2025 Lab model card
45 GLM-4.1V-9B-Thinking 54.4% Jul 2025 Lab model card
46 Nemotron Nano 12B v2 VL 53.6% Nov 2025 Lab model card
47 Gemma 4 E2B 52.4% Apr 2026 Lab model card
48 Nemotron 3.5 Bolt E4B 46.3% Oct 2026 Lab model card
49 Kimi k1.5 38.6% Jan 2025 Lab model card
50 Qwen2.5-VL-72B 38.1% Jan 2025 Lab model card
51 Kimi-VL-A3B-Thinking 36.8% Apr 2025 Lab model card
52 Phi-4-Reasoning-Vision-15B 36.2% Mar 2026 Lab model card
53 Qwen2.5-Omni-7B 25% Mar 2025 Lab model card
#ModelScoreSettingDateSource
1 GPT-5.4 96.1% xhigh, with python Mar 2026 Official leaderboard
2 Gemini 3.1 Pro 95.7% thinking high, with python Feb 2026 Official leaderboard
3 Kimi K2.6 93.2% with python Apr 2026 Official leaderboard
4 Gemini 3 Flash 90.6% Dec 2025 Official leaderboard
5 Seed2.0 Lite (0428) 89.8% Apr 2026 Official leaderboard
6 Seed2.0 Pro 88.8% Feb 2026 Official leaderboard
7 Qwen3.5 397B-A17B 88.6% Feb 2026 Official leaderboard
8 Gemini 3 Pro 86.6% Nov 2025 Official leaderboard
9 Seed2.0 Lite (0215) 86.4% Feb 2026 Official leaderboard
10 Qwen3.5-122B-A10B 86.2% Feb 2026 Official leaderboard
11 Qwen3.5-27B 86% Feb 2026 Official leaderboard
12 Gemma 4 31B IT 85.6% Apr 2026 Official leaderboard
13 Kimi K2.5 85% thinking, with python Jan 2026 Official leaderboard
14 Claude Opus 4.6 84.6% max, with python Feb 2026 Official leaderboard
15 Qwen3.5-35B-A3B 83.9% Feb 2026 Official leaderboard
16 GPT-5.2 83% Dec 2025 Official leaderboard
17 Gemma 4 26B A4B 82.4% Apr 2026 Official leaderboard
18 Seed1.8 81.3% Dec 2025 Official leaderboard
19 Seed1.6-Thinking 77.2% Jun 2025 Official leaderboard
20 Claude Opus 4.5 77.1% Nov 2025 Official leaderboard
21 Qwen3-VL-235B-A22B 74.6% thinking Sep 2025 Official leaderboard
22 Gemini 2.5 Pro Mar 2025 73.3% Mar 2025 Official leaderboard
23 GPT-5 72% Aug 2025 Official leaderboard
24 GPT-5 mini 71.9% Aug 2025 Official leaderboard
25 Claude Sonnet 4.5 71.1% Sep 2025 Official leaderboard
26 dots.vlm1 69.6% Aug 2025 Official leaderboard
27 Seed1.5-VL 68.7% May 2025 Official leaderboard
28 Claude Opus 4.1 66% thinking Aug 2025 Official leaderboard
29 Qwen3-VL-235B-A22B-Instruct 66% Sep 2025 Official leaderboard
30 GLM-4.5V 65.6% Aug 2025 Official leaderboard
31 Step 3 64.8% Jul 2025 Official leaderboard
32 InternVL3.5-241B-A28B 63.9% Aug 2025 Official leaderboard
33 GLM-4.6V 63.5% Dec 2025 Official leaderboard
34 MiMo-VL-7B-RL 60.4% May 2025 Official leaderboard
35 o1 60.3% Dec 2024 Official leaderboard
36 MiMo-VL-7B-RL-2508 59.6% Aug 2025 Official leaderboard
37 Qwen3-VL-8B-Thinking 59.6% Oct 2025 Official leaderboard
38 Gemma 4 E4B 59.5% Mar 2026 Official leaderboard
39 Claude 3.7 Sonnet 58.6% Feb 2025 Official leaderboard
40 o4-mini 58% Apr 2025 Official leaderboard
41 MiMo-VL-7B-SFT 57.9% May 2025 Official leaderboard
42 Kimi-VL-A3B-Thinking-2506 56.9% Jun 2025 Official leaderboard
43 InternVL3.5-30B-A3B 55.7% Aug 2025 Official leaderboard
44 GLM-4.1V-9B-Thinking 54.4% Jul 2025 Official leaderboard
45 GLM-4.6V-Flash 54% Dec 2025 Official leaderboard
46 InternVL3.5-38B 54% Aug 2025 Official leaderboard
47 Ovis2.5-9B 53.9% Aug 2025 Official leaderboard
48 Skywork-R1V3-38B 52.6% Jul 2025 Official leaderboard
49 Gemma 4 E2B 52.4% Mar 2026 Official leaderboard
50 InternVL3.5-8B 52% Aug 2025 Official leaderboard
51 Qianfan-VL-70B 50.3% Sep 2025 Official leaderboard
52 Skywork-R1V2-38B 49.7% Apr 2025 Official leaderboard
53 Doubao-1.5-pro 48.6% Jan 2025 Official leaderboard
54 Gemini 2.0 Pro 48.1% Feb 2025 Official leaderboard
55 GPT-4.5 47.3% Feb 2025 Official leaderboard
56 Gemma 3 27B 46% non-thinking Mar 2025 Official leaderboard
57 Keye-VL-8B-Preview 46% Jun 2025 Official leaderboard
58 InternVL3-78B 43.1% Apr 2025 Official leaderboard
59 Gemini 2.0 Flash Feb 2025 41.3% Feb 2025 Official leaderboard
60 InternVL3.5-4B 40% Aug 2025 Official leaderboard
61 Kimi k1.5 38.6% Jan 2025 Official leaderboard
62 Qwen2.5-VL-72B 38.1% Jan 2025 Official leaderboard
63 Claude 3.5 Sonnet 38% Jun 2024 Official leaderboard
64 Ovis2.5-2B 37.4% Aug 2025 Official leaderboard
65 InternVL3-14B 37% Apr 2025 Official leaderboard
66 Kimi-VL-A3B-Thinking 36.8% Apr 2025 Official leaderboard
67 QVQ-72B-Preview 35.9% Dec 2024 Official leaderboard
68 InternVL3-38B 34.5% Apr 2025 Official leaderboard
69 Qianfan-VL-8B 32.8% Sep 2025 Official leaderboard
70 InternVL2_5-78B 32.2% Dec 2024 Official leaderboard
71 Ovis2-34B 31.9% Feb 2025 Official leaderboard
72 InternVL2.5-38B 31.8% Dec 2024 Official leaderboard
73 GPT-4o May 2024 30.4% May 2024 Official leaderboard
74 GPT-4 Turbo Apr 2024 30.3% Apr 2024 Official leaderboard
75 Ovis2-16B 30.1% Feb 2025 Official leaderboard
76 InternVL3-8B 29% Apr 2025 Official leaderboard
77 InternVL3-9B 27.6% Apr 2025 Official leaderboard
78 Claude 3 Opus 27.1% Feb 2024 Official leaderboard
79 Qwen2-VL-72B-Instruct 25.9% Aug 2024 Official leaderboard
80 LLaVA-OneVision 72B 25.3% Aug 2024 Official leaderboard
81 Qwen2.5-VL-7B 25.1% Jan 2025 Official leaderboard
82 MiniCPM-V-2_6 23.4% Aug 2024 Official leaderboard
83 InternVL2.5-26B 23.1% Dec 2024 Official leaderboard
84 GPT-4V 22.8% Nov 2023 Official leaderboard
85 MiniCPM-o-2_6 21.7% Jan 2025 Official leaderboard
86 InternVL3-2B 21.7% Apr 2025 Official leaderboard
87 Qwen2.5-VL-3B 21.2% Jan 2025 Official leaderboard
88 InternVL2.5-8B 19.7% Dec 2024 Official leaderboard
89 Gemini 1.5 Pro May 2024 19.2% May 2024 Official leaderboard
90 InternVL3-1B 18.8% Apr 2025 Official leaderboard
91 Ovis1.6-Gemma2-9B 18.8% Sep 2024 Official leaderboard
92 Aquila-VL-2B 17.9% Oct 2024 Official leaderboard
93 Gemini 1.0 Pro 17.7% Dec 2023 Official leaderboard
94 InternVL-Chat-V1-2-Plus 17% Feb 2024 Official leaderboard
95 Qwen2-VL-7B-Instruct 16.3% Aug 2024 Official leaderboard
96 Qwen-VL-Max 15.6% Jan 2024 Official leaderboard
97 InternVL2.5-2B 14.7% Dec 2024 Official leaderboard
98 InternLM-XComposer2-VL-7B 14.5% Jan 2024 Official leaderboard
99 InternVL2.5-1B 14.4% Dec 2024 Official leaderboard
100 Qwen2-VL-2B-Instruct 12.4% Aug 2024 Official leaderboard
101 ShareGPT4V-13B 11.9% Dec 2023 Official leaderboard
102 LLaVA-v1.5-13B 11.1% Oct 2023 Official leaderboard
103 Qwen-VL-Plus 10.7% Jan 2024 Official leaderboard
104 ShareGPT4V-7B 10.5% Nov 2023 Official leaderboard
105 llava-v1.5-7b 8.52% Oct 2023 Official leaderboard
#ModelScoreSettingDateSource
1 Qwen 3.8 Max 97.7% Aug 2026 Lab model card
2 LongCat-2.5 96.5% Nov 2026 Lab model card
3 Qwen3.8-Omni-Flash 96.2% Sep 2026 Lab model card
4 GPT-5.4 96.1% xhigh, with python Mar 2026 Official leaderboard
5 Gemini 3.1 Pro 95.7% thinking high, with python Feb 2026 Official leaderboard
6 Qwen3.8 Flash 95.7% Aug 2026 Lab model card
7 Qwen 3.8 27B 94.6% Aug 2026 Lab model card
8 Seed2.1 Pro 94.5% Jun 2026 Lab model card
9 Kimi K3 94.3% Jul 2026 Lab model card
10 Kimi K2.6 93.2% with python Apr 2026 Official leaderboard
11 Inkling 2 Small 92.8% Nov 2026 Lab model card
12 Gemini 3 Flash 90.6% Dec 2025 Official leaderboard
13 Qwen3.7-Plus 90.3% Jun 2026 Lab model card
14 Seed2.0 Lite (0428) 89.8% Apr 2026 Official leaderboard
15 Inkling 2 Mini 89% Nov 2026 Lab model card
16 Seed2.0 Pro 88.8% Feb 2026 Official leaderboard
17 Qwen3.5 397B-A17B 88.6% Feb 2026 Official leaderboard
18 Qwen 3.6 Plus 88% Apr 2026 Lab model card
19 Gemini 3 Pro 86.6% Nov 2025 Official leaderboard
20 Seed2.0 Lite (0215) 86.4% Feb 2026 Official leaderboard
21 Qwen3.5-122B-A10B 86.2% Feb 2026 Official leaderboard
22 Qwen3.5-27B 86% Feb 2026 Official leaderboard
23 Gemma 4 31B IT 85.6% Apr 2026 Official leaderboard
24 Kimi K2.5 85% thinking, with python Jan 2026 Official leaderboard
25 Claude Opus 4.6 84.6% max, with python Feb 2026 Official leaderboard
26 Qwen3.5-35B-A3B 83.9% Feb 2026 Official leaderboard
27 GPT-5.2 83% Dec 2025 Official leaderboard
28 Gemma 4 26B A4B 82.4% Apr 2026 Official leaderboard
29 Seed1.8 81.3% Dec 2025 Official leaderboard
30 Gemma 4 12B 79.7% Jun 2026 Lab model card
31 Qwen3.5-9B 78.9% Mar 2026 Lab model card
32 Nemotron 3.5 Super VL 78.2% Oct 2026 Lab model card
33 Seed1.6-Thinking 77.2% Jun 2025 Official leaderboard
34 Claude Opus 4.5 77.1% Nov 2025 Official leaderboard
35 EXAONE 4.5 33B 75.2% Apr 2026 Lab model card
36 Qwen3-VL-235B-A22B 74.6% thinking Sep 2025 Official leaderboard
37 Qwen3.5-4B 74.6% Mar 2026 Lab model card
38 ERNIE 5.0 74.3% Feb 2026 Lab model card
39 MiniCPM-V-4.7-35B-A3B 74.2% Oct 2026 Lab model card
40 Gemini 2.5 Pro Mar 2025 73.3% Mar 2025 Official leaderboard
41 Qwen3.5-Omni-Plus 73% Mar 2026 Lab model card
42 GPT-5 72% Aug 2025 Official leaderboard
43 GPT-5 mini 71.9% Aug 2025 Official leaderboard
44 Claude Sonnet 4.5 71.1% Sep 2025 Official leaderboard
45 STEP3-VL-10B 70.8% Jan 2026 Lab model card
46 DiffusionGemma 26B A4B 70.5% Jun 2026 Lab model card
47 dots.vlm1 69.6% Aug 2025 Official leaderboard
48 Seed1.5-VL 68.7% May 2025 Official leaderboard
49 Claude Opus 4.1 66% thinking Aug 2025 Official leaderboard
50 Qwen3-VL-235B-A22B-Instruct 66% Sep 2025 Official leaderboard
51 GLM-4.5V 65.6% Aug 2025 Official leaderboard
52 Step 3 64.8% Jul 2025 Official leaderboard
53 InternVL3.5-241B-A28B 63.9% Aug 2025 Official leaderboard
54 GLM-4.6V 63.5% Dec 2025 Official leaderboard
55 Intern-S1 62.5% Jul 2025 Lab model card
56 ERNIE-4.5-VL-28B-A3B-Thinking 61.9% Nov 2025 Lab model card
57 MiMo-VL-7B-RL 60.4% May 2025 Official leaderboard
58 o1 60.3% Dec 2024 Official leaderboard
59 MiMo-VL-7B-RL-2508 59.6% Aug 2025 Official leaderboard
60 Qwen3-VL-8B-Thinking 59.6% Oct 2025 Official leaderboard
61 Gemma 4 E4B 59.5% Mar 2026 Official leaderboard
62 Claude 3.7 Sonnet 58.6% Feb 2025 Official leaderboard
63 o4-mini 58% Apr 2025 Official leaderboard
64 MiMo-VL-7B-SFT 57.9% May 2025 Official leaderboard
65 Kimi-VL-A3B-Thinking-2506 56.9% Jun 2025 Official leaderboard
66 Qwen3-Omni-30B-A3B 56.3% Sep 2025 Lab model card
67 InternVL3.5-30B-A3B 55.7% Aug 2025 Official leaderboard
68 GLM-4.1V-9B-Thinking 54.4% Jul 2025 Official leaderboard
69 GLM-4.6V-Flash 54% Dec 2025 Official leaderboard
70 InternVL3.5-38B 54% Aug 2025 Official leaderboard
71 Ovis2.5-9B 53.9% Aug 2025 Official leaderboard
72 Nemotron Nano 12B v2 VL 53.6% Nov 2025 Lab model card
73 Skywork-R1V3-38B 52.6% Jul 2025 Official leaderboard
74 Gemma 4 E2B 52.4% Mar 2026 Official leaderboard
75 InternVL3.5-8B 52% Aug 2025 Official leaderboard
76 Qianfan-VL-70B 50.3% Sep 2025 Official leaderboard
77 Skywork-R1V2-38B 49.7% Apr 2025 Official leaderboard
78 Doubao-1.5-pro 48.6% Jan 2025 Official leaderboard
79 Gemini 2.0 Pro 48.1% Feb 2025 Official leaderboard
80 GPT-4.5 47.3% Feb 2025 Official leaderboard
81 Nemotron 3.5 Bolt E4B 46.3% Oct 2026 Lab model card
82 Gemma 3 27B 46% non-thinking Mar 2025 Official leaderboard
83 Keye-VL-8B-Preview 46% Jun 2025 Official leaderboard
84 InternVL3-78B 43.1% Apr 2025 Official leaderboard
85 Gemini 2.0 Flash Feb 2025 41.3% Feb 2025 Official leaderboard
86 InternVL3.5-4B 40% Aug 2025 Official leaderboard
87 Kimi k1.5 38.6% Jan 2025 Official leaderboard
88 Qwen2.5-VL-72B 38.1% Jan 2025 Official leaderboard
89 Claude 3.5 Sonnet 38% Jun 2024 Official leaderboard
90 Ovis2.5-2B 37.4% Aug 2025 Official leaderboard
91 InternVL3-14B 37% Apr 2025 Official leaderboard
92 Kimi-VL-A3B-Thinking 36.8% Apr 2025 Official leaderboard
93 Phi-4-Reasoning-Vision-15B 36.2% Mar 2026 Lab model card
94 QVQ-72B-Preview 35.9% Dec 2024 Official leaderboard
95 InternVL3-38B 34.5% Apr 2025 Official leaderboard
96 Qianfan-VL-8B 32.8% Sep 2025 Official leaderboard
97 InternVL2_5-78B 32.2% Dec 2024 Official leaderboard
98 Ovis2-34B 31.9% Feb 2025 Official leaderboard
99 InternVL2.5-38B 31.8% Dec 2024 Official leaderboard
100 GPT-4o May 2024 30.4% May 2024 Official leaderboard
101 GPT-4 Turbo Apr 2024 30.3% Apr 2024 Official leaderboard
102 Ovis2-16B 30.1% Feb 2025 Official leaderboard
103 InternVL3-8B 29% Apr 2025 Official leaderboard
104 InternVL3-9B 27.6% Apr 2025 Official leaderboard
105 Claude 3 Opus 27.1% Feb 2024 Official leaderboard
106 Qwen2-VL-72B-Instruct 25.9% Aug 2024 Official leaderboard
107 LLaVA-OneVision 72B 25.3% Aug 2024 Official leaderboard
108 Qwen2.5-VL-7B 25.1% Jan 2025 Official leaderboard
109 Qwen2.5-Omni-7B 25% Mar 2025 Lab model card
110 MiniCPM-V-2_6 23.4% Aug 2024 Official leaderboard
111 InternVL2.5-26B 23.1% Dec 2024 Official leaderboard
112 GPT-4V 22.8% Nov 2023 Official leaderboard
113 MiniCPM-o-2_6 21.7% Jan 2025 Official leaderboard
114 InternVL3-2B 21.7% Apr 2025 Official leaderboard
115 Qwen2.5-VL-3B 21.2% Jan 2025 Official leaderboard
116 InternVL2.5-8B 19.7% Dec 2024 Official leaderboard
117 Gemini 1.5 Pro May 2024 19.2% May 2024 Official leaderboard
118 InternVL3-1B 18.8% Apr 2025 Official leaderboard
119 Ovis1.6-Gemma2-9B 18.8% Sep 2024 Official leaderboard
120 Aquila-VL-2B 17.9% Oct 2024 Official leaderboard
121 Gemini 1.0 Pro 17.7% Dec 2023 Official leaderboard
122 InternVL-Chat-V1-2-Plus 17% Feb 2024 Official leaderboard
123 Qwen2-VL-7B-Instruct 16.3% Aug 2024 Official leaderboard
124 Qwen-VL-Max 15.6% Jan 2024 Official leaderboard
125 InternVL2.5-2B 14.7% Dec 2024 Official leaderboard
126 InternLM-XComposer2-VL-7B 14.5% Jan 2024 Official leaderboard
127 InternVL2.5-1B 14.4% Dec 2024 Official leaderboard
128 Qwen2-VL-2B-Instruct 12.4% Aug 2024 Official leaderboard
129 ShareGPT4V-13B 11.9% Dec 2023 Official leaderboard
130 LLaVA-v1.5-13B 11.1% Oct 2023 Official leaderboard
131 Qwen-VL-Plus 10.7% Jan 2024 Official leaderboard
132 ShareGPT4V-7B 10.5% Nov 2023 Official leaderboard
133 llava-v1.5-7b 8.52% Oct 2023 Official leaderboard

The benchmark's own leaderboard has 153 entries, 105 of them models we track. Switch the scores above to rank them together.

Rankings by part

We read 16 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Angle, Area and Algebra. Skywork-R1V2-38B leads 8 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Angle

16 models we track have a score here.

  1. 1 Skywork-R1V2-38B ↑51 63.2%
  2. 2 Doubao-1.5-pro ↑51 62.4%
  3. 3 GPT-4 Turbo Apr 2024 ↑71 30.6%
  4. 4 InternVL-Chat-V1-2-Plus ↑90 22.5%
  5. 5 GPT-4V ↑79 22%
  6. 6 Ovis1.6-Gemma2-9B ↑85 20.8%
  7. 7 InternLM-XComposer2-VL-7B ↑91 19.7%
  8. 8 Gemini 1.0 Pro ↑85 19.1%
  9. 9 GPT-4o May 2024 ↑64 17.3%
  10. 10 ShareGPT4V-7B ↑94 16.2%

More effort adds a median 6 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-5.4
xhigh
96.1%
Gemini 3.1 Pro
thinking hi… thinking hi…
95.7%
Kimi K2.6
no tools with python
93.2%
Kimi K2.5
thinking, w…
85%
Claude Opus 4.6
max max, with p…
84.6%
GPT-5
minimal
72%

Source: Official MATH-Vision (MathVision) leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5.4 xhigh, with python 96.1%
Gemini 3.1 Pro thinking high, with python 95.7%
Kimi K2.6 with python 93.2%
GPT-5.4 xhigh 92%
Gemini 3 Flash 90.6%
Gemini 3.1 Pro thinking high 89.8%
Seed2.0 Lite (0428) 89.8%
Seed2.0 Pro 88.8%
Qwen3.5 397B-A17B 88.6%
Kimi K2.6 no tools 87.4%
Gemini 3 Pro 86.6%
Seed2.0 Lite (0215) 86.4%
Qwen3.5-122B-A10B 86.2%
Qwen3.5-27B 86%
Gemma 4 31B IT 85.6%
Kimi K2.5 thinking, with python 85%
Claude Opus 4.6 max, with python 84.6%
Kimi K2.5 84.2%
Qwen3.5-35B-A3B 83.9%
GPT-5.2 83%
Gemma 4 26B A4B 82.4%
Seed1.8 81.3%
Seed1.6-Thinking 77.2%
Claude Opus 4.5 77.1%
Qwen3-VL-235B-A22B thinking 74.6%
Gemini 2.5 Pro Mar 2025 73.3%
GPT-5 72%
GPT-5 mini 71.9%
Claude Opus 4.6 max 71.2%
Claude Sonnet 4.5 71.1%
dots.vlm1 69.6%
Seed1.5-VL 68.7%
Claude Opus 4.1 thinking 66%
Qwen3-VL-235B-A22B-Instruct 66%
GLM-4.5V 65.6%
Step 3 64.8%
InternVL3.5-241B-A28B 63.9%
GLM-4.6V 63.5%
MiMo-VL-7B-RL 60.4%
o1 60.3%
MiMo-VL-7B-RL-2508 59.6%
Qwen3-VL-8B-Thinking 59.6%
Gemma 4 E4B 59.5%
Claude 3.7 Sonnet 58.6%
o4-mini 58%
MiMo-VL-7B-SFT 57.9%
Kimi-VL-A3B-Thinking-2506 56.9%
InternVL3.5-30B-A3B 55.7%
GLM-4.1V-9B-Thinking 54.4%
GLM-4.6V-Flash 54%
InternVL3.5-38B 54%
Ovis2.5-9B 53.9%
Skywork-R1V3-38B 52.6%
Gemma 4 E2B 52.4%
InternVL3.5-8B 52%
Qianfan-VL-70B 50.3%
Skywork-R1V2-38B 49.7%
Doubao-1.5-pro 48.6%
Gemini 2.0 Pro 48.1%
GPT-4.5 47.3%
Gemma 3 27B non-thinking 46%
Keye-VL-8B-Preview 46%
GPT-5 minimal 45.8%
InternVL3-78B 43.1%
Gemini 2.0 Flash Feb 2025 41.3%
InternVL3.5-4B 40%
Kimi k1.5 38.6%
Qwen2.5-VL-72B 38.1%
Claude 3.5 Sonnet 38%
Ovis2.5-2B 37.4%
InternVL3-14B 37%
Kimi-VL-A3B-Thinking 36.8%
QVQ-72B-Preview 35.9%
InternVL3-38B 34.5%
Qianfan-VL-8B 32.8%
InternVL2_5-78B 32.2%
Ovis2-34B 31.9%
InternVL2.5-38B 31.8%
GPT-4o May 2024 30.4%
GPT-4 Turbo Apr 2024 30.3%
Ovis2-16B 30.1%
InternVL3-8B 29%
InternVL3-9B 27.6%
Claude 3 Opus 27.1%
Qwen2-VL-72B-Instruct 25.9%
LLaVA-OneVision 72B 25.3%
Qwen2.5-VL-7B 25.1%
MiniCPM-V-2_6 23.4%
InternVL2.5-26B 23.1%
GPT-4V 22.8%
MiniCPM-o-2_6 21.7%
InternVL3-2B 21.7%
Qwen2.5-VL-3B 21.2%
InternVL2.5-8B 19.7%
Gemini 1.5 Pro May 2024 19.2%
InternVL3-1B 18.8%
Ovis1.6-Gemma2-9B 18.8%
Aquila-VL-2B 17.9%
Gemini 1.0 Pro 17.7%
InternVL-Chat-V1-2-Plus 17%
Qwen2-VL-7B-Instruct 16.3%
Qwen-VL-Max 15.6%
InternVL2.5-2B 14.7%
InternLM-XComposer2-VL-7B 14.5%
InternVL2.5-1B 14.4%
Qwen2-VL-2B-Instruct 12.4%
ShareGPT4V-13B 11.9%
LLaVA-v1.5-13B 11.1%
Qwen-VL-Plus 10.7%
ShareGPT4V-7B 10.5%
llava-v1.5-7b 8.52%

11 models have held number one in labs’ own numbers since its release in February 2024 12 models have held number one across all independent runs since its release in February 2024 13 models have held number one with claims included since its release in February 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 and what labs claim in their cards.

OpenAI Anthropic Google DeepMind ByteDance Seed Moonshot AI Qwen
Own leaderboard Passed 85% in Nov 2025, by Gemini 3 Pro
GPT-5.4 96.1%
Lab claims Passed 85% in Feb 2026, by Seed2.0 Pro
Qwen 3.8 Max 97.7%

Sources: Official MATH-Vision (MathVision) leaderboard and labs’ model cards

Gemini 3 Pro on the official leaderboard passed 85% in November 2025

That was 21 months after its release. By the same age, 34% of the benchmarks from 2024 had passed 85%. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 2027 Near the ceiling (85%) Qwen 3.8 Max Gemini 3 Flash Seed1.6-Thinking o1 Claude 3.5 Sonnet GPT-4o GPT-4V 0% 20% 40% 60% 80% 100% 2024 2025 2026 2027 Near the ceiling (85%) Qwen 3.8 Max Gemini 3 Flash

Sources: Official MATH-Vision (MathVision) leaderboard and labs’ model cards

8 labs put it in a launch post in the last six months

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.

Moonshot AI
Qwen
ByteDance Seed
Shanghai AI Lab
StepFun
Z.ai
Baidu
Microsoft
Google DeepMind
LG AI Research
OpenBMB (ModelBest)
NVIDIA
Meituan
Thinking Machines
DateLaunch postScoreAs the post writes it
Nov 5, 2026 Inkling 2 Small 92.8 MathVision, 92.8%
Nov 2, 2026 LongCat-2.5 96.5 MathVision (w/ tools)
Oct 15, 2026 Nemotron 3.5 Super VL 78.2 MathVision
Oct 7, 2026 MiniCPM-V-4.7-35B-A3B 74.2 MathVision
Sep 18, 2026 Qwen3.8-Omni-Flash 96.2 MathVision Without CI 91.8 With CI 96.2
Aug 26, 2026 Qwen3.8 Flash 95.7 Visual math problem solving MathVision Without CI 90.6 With CI 95.7
Aug 14, 2026 Qwen 3.8 27B 94.6 Visual math problem solving MathVision Without CI 90.0 With CI 94.6
Aug 3, 2026 Qwen 3.8 Max 97.7 MathVision (without CI / with CI) 95.2 / 97.7
Jul 16, 2026 Kimi K3 94.3 MathVision (without python)
Jun 23, 2026 Seed2.1 Pro 94.5 MathVision (w. Tool) 92.6 (94.5)
Jun 10, 2026 DiffusionGemma 26B A4B 70.5 MATH-Vision
Jun 3, 2026 Gemma 4 12B 79.7 MATH-Vision
Jun 1, 2026 Qwen3.7-Plus 90.3 MathVision
Apr 28, 2026 Seed2.0 Lite (0428) 89.8 MathVision
Apr 20, 2026 Kimi K2.6 87.4 MathVision
Apr 9, 2026 EXAONE 4.5 33B 75.2 MathVision
Apr 2, 2026 Gemma 4 E4B 59.5 MATH-Vision
Apr 2, 2026 Gemma 4 E2B 52.4 MATH-Vision
Apr 2, 2026 Gemma 4 26B A4B 82.4 MATH-Vision
Apr 2, 2026 Qwen 3.6 Plus 88 MathVision
Apr 2, 2026 Gemma 4 31B IT 85.6 MATH-Vision
Mar 30, 2026 Qwen3.5-Omni-Plus 73 MathVision
Mar 4, 2026 Phi-4-Reasoning-Vision-15B 36.2 MathVision_MINI (default mode)
Mar 2, 2026 Qwen3.5-4B 74.6 MathVision
Mar 2, 2026 Qwen3.5-9B 78.9 MathVision
Feb 24, 2026 Qwen3.5-35B-A3B 83.9 MathVision
Feb 24, 2026 Qwen3.5-27B 86 MathVision
Feb 24, 2026 Qwen3.5-122B-A10B 86.2 MathVision
Feb 16, 2026 Qwen3.5 397B-A17B 88.6 MathVision
Feb 14, 2026 Seed2.0 Pro 88.8 MathVision
Feb 6, 2026 ERNIE 5.0 74.3 MathVision
Jan 27, 2026 Kimi K2.5 84.2 MathVision
Jan 13, 2026 STEP3-VL-10B 70.8 MathVision (SeRe)
Dec 18, 2025 Seed1.8 81.3 MathVision
Nov 11, 2025 ERNIE-4.5-VL-28B-A3B-Thinking 61.9 MathVision
Sep 23, 2025 Qwen3-VL-235B-A22B 74.6 MathVision
Sep 22, 2025 Qwen3-Omni-30B-A3B 56.3 MathVision_full
Aug 25, 2025 InternVL3.5-241B-A28B 63.9 MathVision
Aug 11, 2025 GLM-4.5V 65.6 MathVision
Jul 31, 2025 Step 3 64.8 MATH-Vision
Jul 24, 2025 Intern-S1 62.5 MathVision
Jun 25, 2025 Seed1.6-Thinking 70.3 MathVision
Jun 21, 2025 Kimi-VL-A3B-Thinking-2506 56.9 MATH-Vision (Pass@1)
May 12, 2025 Seed1.5-VL 68.7 MathVision
Apr 10, 2025 Kimi-VL-A3B-Thinking 36.8 MathVision (full) (Pass@1)
Mar 27, 2025 Qwen2.5-Omni-7B 25 MathVisionfull
Jan 28, 2025 Qwen2.5-VL-72B 38.1 MathVision_FULL
Jan 20, 2025 Kimi k1.5 38.6 MathVision-Full (Pass@1), long-CoT

32 system cards report it, from 12 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
Nov 20, 2026 Inkling 2 Mini model card 89 MathVision Capabilities
Oct 23, 2026 Nemotron 3.5 Bolt E4B technical report 46.3 MathVision, reasoning, 43.72 non-reasoning (Table 8) Capabilities
Oct 15, 2026 Nemotron 3.5 Super VL technical report 78.2 MathVision (Table 10) Capabilities
Sep 22, 2026 Qwen3.8-Omni-Flash technical report 91.8 MathVision without CI, some ground truth corrected (Table 3). With CI 96.2 Capabilities
Jul 27, 2026 Kimi K3 technical report 94.3 Math-Vision without Python, cell 94.3 / 97.8 (Table 2) Capabilities
Jul 24, 2026 Gemma 4 E4B technical report 59.5 MATH-Vision, max resolution, thinking (Table 6) Capabilities
Jul 24, 2026 Gemma 4 E2B technical report 52.4 MATH-Vision, max resolution, thinking (Table 6) Capabilities
Jul 24, 2026 Gemma 4 31B IT technical report 85.6 MATH-Vision, thinking, max resolution (Table 6) Capabilities
Jul 24, 2026 Gemma 4 26B A4B technical report 82.4 MATH-Vision, thinking, max resolution (Table 6) Capabilities
Jun 23, 2026 Seed2.1 Pro model card 92.6 MathVision, no tools, 94.5 with tools (Table 9) Capabilities
Apr 17, 2026 Qwen3.5-Omni-Plus technical report 73 MathVision (Table 6) Capabilities
Apr 9, 2026 EXAONE 4.5 33B technical report 75.2 MATH-Vision (Table 2, Reasoning mode) Capabilities
Mar 21, 2026 Seed1.8 model card 81.3 MathVision (Table 2) Capabilities
Mar 4, 2026 Phi-4-Reasoning-Vision-15B technical report 36.2 MathVision MINI, default mode, 34.2 force nothink, 36.2 force thinking (Tables 5 and 6) Capabilities
Feb 14, 2026 Seed2.0 Pro model card 88.8 MathVision (Table 8) Capabilities
Feb 4, 2026 ERNIE 5.0 technical report 74.3 MathVision (Table 3) Capabilities
Jan 30, 2026 Kimi K2.5 technical report 84.2 MathVision (Table 4) Capabilities
Jan 14, 2026 STEP3-VL-10B technical report 70.8 MathVision, SeRe (Tables 1 and 3). PaCoRe 75.95 Capabilities
Jan 1, 2026 GLM-4.6V technical report 63.5 MathVision, GLM-4.6V thinking (Table 2) Capabilities
Nov 26, 2025 Qwen3-VL-235B-A22B technical report 74.6 MathVision (Table 2) Capabilities
Nov 6, 2025 Nemotron Nano 12B v2 VL technical report 53.6 MathVision, reasoning on (Table 3). Reasoning off 31.5 Capabilities
Sep 22, 2025 Qwen3-Omni-30B-A3B technical report 56.3 MATH-Vision full, Instruct (Table 9). Thinking 62.9 (Table 10) Capabilities
Aug 25, 2025 InternVL3.5-241B-A28B technical report 63.9 MathVision (Table 2, Table 3). With parallel thinking 65.9 Capabilities
Aug 21, 2025 Intern-S1 technical report 62.5 MathVision (Table 2) Capabilities
Aug 13, 2025 GLM-4.5V technical report 65.6 MathVision, GLM-4.5V thinking (Table 2) Capabilities
Jul 1, 2025 GLM-4.1V-9B-Thinking technical report 54.4 MathVision, GLM-4.1V-9B-Thinking thinking (Table 2) Capabilities
Jun 23, 2025 Kimi-VL-A3B-Thinking-2506 technical report 56.9 MathVision (full) (Pass@1) (Table 4) Capabilities
May 11, 2025 Seed1.5-VL technical report 68.7 MathVision, thinking (Table 6). Non-thinking 65.5 Capabilities
Apr 10, 2025 Kimi-VL-A3B-Thinking technical report 36.8 MathVision (full) (Pass@1) (Table 4). Figure 13 shows 18.7 at 1k rising to 36.8 at 16k max thinking tokens Capabilities
Mar 26, 2025 Qwen2.5-Omni-7B technical report 25 MathVision full (Table 5) Capabilities
Feb 19, 2025 Qwen2.5-VL-72B technical report 38.1 MATH-Vision full (Table 3) Capabilities
Jan 20, 2025 Kimi k1.5 technical report 38.6 MathVision-Full (Pass@1), long-CoT (Table 2). Short-CoT 31.0 (Table 3) Capabilities

The lab's number comes out ahead in 0 of 3 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. 22 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
Seed1.6-Thinking 70.3% MathVision 77.2% Leaderboard −6.9 pts
Kimi K2.6 87.4% MathVision 93.2% Leaderboard, with python −5.8 pts
Kimi K2.5 84.2% MathVision 85% Leaderboard, thinking, with python −0.8 pts
Kimi k1.5 38.6% MathVision-Full (Pass@1), long-CoT 38.6% Leaderboard Same
Qwen2.5-VL-72B 38.1% MathVision_FULL 38.1% Leaderboard Same
Kimi-VL-A3B-Thinking 36.8% MathVision (full) (Pass@1) 36.8% Leaderboard Same
Seed1.5-VL 68.7% MathVision 68.7% Leaderboard Same
Kimi-VL-A3B-Thinking-2506 56.9% MATH-Vision (Pass@1) 56.9% Leaderboard Same
Step 3 64.8% MATH-Vision 64.8% Leaderboard Same
GLM-4.5V 65.6% MathVision 65.6% Leaderboard Same
InternVL3.5-241B-A28B 63.9% MathVision 63.9% Leaderboard Same
Qwen3-VL-235B-A22B 74.6% MathVision 74.6% Leaderboard, thinking Same
Seed1.8 81.3% MathVision 81.3% Leaderboard Same
Seed2.0 Pro 88.8% MathVision 88.8% Leaderboard Same
Qwen3.5 397B-A17B 88.6% MathVision 88.6% Leaderboard Same
Qwen3.5-122B-A10B 86.2% MathVision 86.2% Leaderboard Same
Qwen3.5-27B 86% MathVision 86% Leaderboard Same
Qwen3.5-35B-A3B 83.9% MathVision 83.9% Leaderboard Same
Gemma 4 31B IT 85.6% MATH-Vision 85.6% Leaderboard Same
Gemma 4 26B A4B 82.4% MATH-Vision 82.4% Leaderboard Same
Gemma 4 E2B 52.4% MATH-Vision 52.4% Leaderboard Same
Gemma 4 E4B 59.5% MATH-Vision 59.5% Leaderboard Same
Seed2.0 Lite (0428) 89.8% MathVision 89.8% Leaderboard Same
GLM-4.1V-9B-Thinking 54.4% MathVision, GLM-4.1V-9B-Thinking thinking (Table 2) 54.4% Leaderboard Same
GLM-4.6V 63.5% MathVision, GLM-4.6V thinking (Table 2) 63.5% Leaderboard Same

More in Vision

All 636 in Vision
  • MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. Qwen 3.6 Plus 86%
  • MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
  • CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
  • MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy. Seed2.1 Pro 90.7%
  • MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy. Qwen 3.8 Max 85.9%
  • ChartQA In use Models answer human-written and generated questions about chart images that need visual and arithmetic reasoning. Scored by relaxed accuracy with a 5 percent numeric tolerance. Command A Vision 90.9%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (153 entries), launch posts (48 scores) and system cards (32 scores).
  • Pass Models scored 133 models we track, plus 48 other entries on its leaderboard.
  • Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is Kimi K3 (#25).
  • Note Headroom Saturated. The best is 97.7%, past 85% since November 2025.
Leaderboard import 3 pass
  • Pass Importer We read the board on mathllm.github.io: 111 rows for models we track and 48 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 2 pass
  • Pass Lab claims All 25 pairs agree within 10 points.
  • Pass Known issues No published audit found problems.

How we check