Benchmark Almanac

ARC-AGI

Claude Fable 5 98.5% leads since June 2026
Claude Fable 5 98.5% leads since June 2026
In use
Version 1

The official leaderboard hasn’t scored a model released after GPT-6.2 Astra (November 2026). Labs report their own numbers for one newer model: Inkling 2 Mini at 84.5%.

Few-shot grid puzzles: infer a hidden transformation from about three examples and produce the output grid. Semi-private sets verified by ARC Prize.

ARC-AGI-1 is saturated at about 97 percent. ARC-AGI-2 and ARC-AGI-3 are its successors in the series.

Released
November 2019
Kind
Abstract puzzles and logic
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
Public, plus a private test set
Measures
Capability
Language
English
In and out
Images → Images
Interaction
Single turn
Run by
A host
Strongest model tested
GPT-6.2 Astra #1 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 19th, 2026

In the authors’ words

To make deliberate progress towards more intelligent and more human-like artificial systems, we need to be following an appropriate feedback signal: we need to be able to define and evaluate intelligence in a way that enables comparisons between two systems, as well as comparisons with humans. Over the past hundred years, there has been an abundance of attempts to define and measure intelligence, across both the fields of psychology and AI. We summarize and critically assess these definitions and evaluation approaches, while making apparent the two historical conceptions of intelligence that have implicitly guided them. We note that in practice, the contemporary AI community still gravitates towards benchmarking intelligence by comparing the skill exhibited by AIs and humans at specific tasks such as board games and video games. We argue that solely measuring skill at any given task falls short of measuring intelligence, because skill is heavily modulated by prior knowledge and experience: unlimited priors or unlimited training data allow experimenters to "buy" arbitrary levels of skills for a system, in a way that masks the system's own generalization power. We then articulate a new formal definition of intelligence based on Algorithmic Information Theory, describing intelligence as skill-acquisition efficiency and highlighting the concepts of scope, generalization difficulty, priors, and experience. Using this definition, we propose a set of guidelines for what a general AI benchmark should look like. Finally, we present a benchmark closely following these guidelines, the Abstraction and Reasoning Corpus (ARC), built upon an explicit set of priors designed to be as close as possible to innate human priors. We argue that ARC can be used to measure a human-like form of general fluid intelligence and that it enables fair general intelligence comparisons between AI systems and humans.

Abstract of On the Measure of Intelligence (arXiv:1911.01547, CC0)
101
Entries on its own leaderboard, 98 models we track
16
Launch posts list it, from 7 labs
16
System cards report it, from 7 labs

9 models share the top score of 98.5% 9 models share the top score of 98.5%

Best score per entry from the benchmark's own leaderboard, read on November 18th, 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. This view shows only models we track, each at its best.

Claude Fable 5 leads with 98.5%, 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.

101 entries. Scroll sideways for the rest.

  1. 98.5 Claude Fable 5
  2. 98.5 Gemini 3.8 Flash
  3. 98.5 GPT-6 Astra
  4. 98.5 Claude Opus 5.5
  5. 98.5 GPT-6.1 Sol
  6. 98.5 Gemini 4 Argon
  7. 98.5 Gemini 4 Carbon
  8. 98.5 Claude Fable 5.2
  9. 98.5 GPT-6.2 Astra
  10. 98.0 Gemini 3.1 Pro
  11. 97.5 GPT-5.6 Sol
  12. 97.5 Claude Opus 5
  13. 97.5 Claude Fable 5.1
  14. 97.5 GPT-6.2 Sol
  15. 96.5 GPT-5.5 Pro
  16. 96.5 GPT-5.6 Terra
  17. 96.5 Claude Sonnet 5.5
  18. 96.5 Inkling 2 Small
  19. 96.0 Gemini 3 Deep Think
  20. 95.5 Gemini 3.7 Flash
  21. 95.5 GPT-6 Sol
  22. 95.0 GPT-5.5
  23. 94.5 GPT-5.2 (Refinement)
  24. 94.5 GPT-5.4 Pro
  25. 94.5 Kimi K3
  26. 94.5 DeepSeek V4.1 Flash
  27. 94.5 DeepSeek V4.1 Flash 1017
  28. 94.0 Claude Opus 4.6
  29. 93.7 GPT-5.4
  30. 93.5 Claude Opus 4.7
  31. 92.5 Gemini 3.5 Flash
  32. 92.5 Claude Opus 4.8
  33. 91.2 Gemini 3.6 Flash
  34. 91.0 GLM-5.3-Flash
  35. 90.7 GPT-5.6 Luna
  36. 90.5 GPT-5.2 Pro
  37. 90.5 DeepSeek V4 Pro 0813
  38. 90.2 Grok 4.7
  39. 89.5 Grok 4.20
  40. 89.0 DeepSeek V4 Flash 0731
  41. 89.0 Claude Haiku 5.5
  42. 89.0 Muse Spark Open
  43. 88.5 Mercury 3
  44. 87.5 Grok 4.6
  45. 87.5 Qwen 3.8 27B
  46. 87.2 Grok 4.5
  47. 86.7 GPT-6 Luna
  48. 86.5 Claude Sonnet 4.6
  49. 86.2 GPT-5.2
  50. 84.7 Gemini 3 Flash
  51. 84.0 Inkling-Small
  52. 80.0 Claude Opus 4.5
  53. 79.6 Grok 4 (Refinement)
  54. 79.5 Inkling
  55. 77.0 GLM-5.2
  56. 75.0 Gemini 3 Pro
  57. 74.0 Beam 1.0
  58. 72.8 GPT-5.1
  59. 70.2 GPT-5 Pro
  60. 66.7 Grok 4
  61. 65.7 GPT-5
  62. 65.3 Kimi K2.5
  63. 63.7 Claude Sonnet 4.5
  64. 63.7 MiniMax-M2.5
  65. 63.7 GPT-5.4 Mini
  66. 60.8 o3
  67. 59.3 o3-pro
  68. 58.7 o4-mini
  69. 57.0 DeepSeek-V3.2
  70. 54.3 GPT-5 mini
  71. 53.5 Gemini 3.5 Flash-Lite
  72. 51.5 GPT-5.4 Nano
  73. 48.5 Grok 4 Fast
  74. 47.7 Claude Haiku 4.5
  75. 44.7 GLM-5
  76. 41.0 Gemini 2.5 Pro
  77. 40.0 Claude Sonnet 4
  78. 40.0 tiny-recursion-model
  79. 35.7 Claude Opus 4
  80. 34.5 o3-mini
  81. 33.3 Gemini 2.5 Flash
  82. 28.6 Claude 3.7 Sonnet
  83. 27.3 codex-mini-2025-05-16
  84. 22.7 Nemotron 3.5 Super VL
  85. 21.2 DeepSeek-R1
  86. 20.7 GPT-5 nano
  87. 16.5 Grok-3 mini
  88. 15.8 DeepSeek-R1
  89. 14.0 o1-mini
  90. 11.0 Qwen3-235B-A22B-Instruct
  91. 10.7 Mercury 3 Preview (Instant)
  92. 10.3 GPT-4.5
  93. 6.12 Magistral Medium 1.0
  94. 5.50 Grok 3
  95. 5.50 GPT-4.1
  96. 5.00 Magistral Small 1.0
  97. 4.50 GPT-4o
  98. 4.38 Llama 4 Maverick
  99. 3.50 GPT-4.1 mini
  100. 0.50 Llama 4 Scout
  101. 0.0 GPT-4.1 nano
  1. 1 Claude Fable 5 max 98.5%
  2. 2 Gemini 3.8 Flash high 98.5%
  3. 3 GPT-6 Astra xhigh 98.5%
  4. 4 Claude Opus 5.5 high 98.5%
  5. 5 GPT-6.1 Sol xhigh 98.5%
  6. 6 Gemini 4 Argon high 98.5%
  7. 7 Gemini 4 Carbon medium 98.5%
  8. 8 Claude Fable 5.2 xhigh 98.5%
  9. 9 GPT-6.2 Astra max 98.5%
  10. 10 Gemini 3.1 Pro 98%
  11. 11 GPT-5.6 Sol xhigh 97.5%
  12. 12 Claude Opus 5 high 97.5%
  13. 13 Claude Fable 5.1 max 97.5%
  14. 14 GPT-6.2 Sol max 97.5%
  15. 15 GPT-5.5 Pro high 96.5%
  16. 16 GPT-5.6 Terra max 96.5%
  17. 17 Claude Sonnet 5.5 max 96.5%
  18. 18 Inkling 2 Small xhigh 96.5%
  19. 19 Gemini 3 Deep Think 96%
  20. 20 Gemini 3.7 Flash high 95.5%
  21. 21 GPT-6 Sol max 95.5%
  22. 22 GPT-5.5 xhigh 95%
  23. 23 GPT-5.2 (Refinement) 94.5%
  24. 24 GPT-5.4 Pro xhigh 94.5%
  25. 25 Kimi K3 max 94.5%
  26. 26 DeepSeek V4.1 Flash max 94.5%
  27. 27 DeepSeek V4.1 Flash 1017 high 94.5%
  28. 28 Claude Opus 4.6 120k, high 94%
  29. 29 GPT-5.4 xhigh 93.7%
  30. 30 Claude Opus 4.7 high 93.5%
  31. 31 Gemini 3.5 Flash high 92.5%
  32. 32 Claude Opus 4.8 max 92.5%
  33. 33 Gemini 3.6 Flash high 91.2%
  34. 34 GLM-5.3-Flash max 91%
  35. 35 GPT-5.6 Luna max 90.7%
  36. 36 GPT-5.2 Pro xhigh 90.5%
  37. 37 DeepSeek V4 Pro 0813 low 90.5%
  38. 38 Grok 4.7 high 90.2%
  39. 39 Grok 4.20 reasoning 89.5%
  40. 40 DeepSeek V4 Flash 0731 max 89%
  41. 41 Claude Haiku 5.5 max 89%
  42. 42 Muse Spark Open xhigh 89%
  43. 43 Mercury 3 high 88.5%
  44. 44 Grok 4.6 medium 87.5%
  45. 45 Qwen 3.8 27B xhigh 87.5%
  46. 46 Grok 4.5 medium 87.2%
  47. 47 GPT-6 Luna max 86.7%
  48. 48 Claude Sonnet 4.6 high 86.5%
  49. 49 GPT-5.2 xhigh 86.2%
  50. 50 Gemini 3 Flash high 84.7%
  51. 51 Inkling-Small xhigh 84%
  52. 52 Claude Opus 4.5 thinking, 64k 80%
  53. 53 Grok 4 (Refinement) 79.6%
  54. 54 Inkling 79.5%
  55. 55 GLM-5.2 77%
  56. 56 Gemini 3 Pro 75%
  57. 57 Beam 1.0 xhigh 74%
  58. 58 GPT-5.1 thinking, high 72.8%
  59. 59 GPT-5 Pro 70.2%
  60. 60 Grok 4 thinking 66.7%
  61. 61 GPT-5 high 65.7%
  62. 62 Kimi K2.5 65.3%
  63. 63 Claude Sonnet 4.5 thinking 32k 63.7%
  64. 64 MiniMax-M2.5 63.7%
  65. 65 GPT-5.4 Mini xhigh 63.7%
  66. 66 o3 high 60.8%
  67. 67 o3-pro high 59.3%
  68. 68 o4-mini high 58.7%
  69. 69 DeepSeek-V3.2 57%
  70. 70 GPT-5 mini high 54.3%
  71. 71 Gemini 3.5 Flash-Lite high 53.5%
  72. 72 GPT-5.4 Nano xhigh 51.5%
  73. 73 Grok 4 Fast 48.5%
  74. 74 Claude Haiku 4.5 thinking 32k 47.7%
  75. 75 GLM-5 44.7%
  76. 76 Gemini 2.5 Pro Jun 2025, thinking 16k 41%
  77. 77 Claude Sonnet 4 thinking 16k 40%
  78. 78 tiny-recursion-model 40%
  79. 79 Claude Opus 4 thinking 16k 35.7%
  80. 80 o3-mini high 34.5%
  81. 81 Gemini 2.5 Flash May 2025 33.3%
  82. 82 Claude 3.7 Sonnet 16k 28.6%
  83. 83 codex-mini-2025-05-16 27.3%
  84. 84 Nemotron 3.5 Super VL high 22.7%
  85. 85 DeepSeek-R1 May 2025 21.2%
  86. 86 GPT-5 nano medium 20.7%
  87. 87 Grok-3 mini low 16.5%
  88. 88 DeepSeek-R1 15.8%
  89. 89 o1-mini 14%
  90. 90 Qwen3-235B-A22B-Instruct Jul 2025 11%
  91. 91 Mercury 3 Preview (Instant) 10.7%
  92. 92 GPT-4.5 10.3%
  93. 93 Magistral Medium 1.0 thinking 6.12%
  94. 94 Grok 3 5.50%
  95. 95 GPT-4.1 5.50%
  96. 96 Magistral Small 1.0 5.00%
  97. 97 GPT-4o Nov 2024 4.50%
  98. 98 Llama 4 Maverick 4.38%
  99. 99 GPT-4.1 mini 3.50%
  100. 100 Llama 4 Scout 0.50%
  101. 101 GPT-4.1 nano 0%

Source: Official ARC-AGI leaderboard

98 models. Scroll sideways for the rest.

  1. 98.5 Claude Fable 5
  2. 98.5 Gemini 3.8 Flash
  3. 98.5 GPT-6 Astra
  4. 98.5 Claude Opus 5.5
  5. 98.5 GPT-6.1 Sol
  6. 98.5 Gemini 4 Argon
  7. 98.5 Gemini 4 Carbon
  8. 98.5 Claude Fable 5.2
  9. 98.5 GPT-6.2 Astra
  10. 98.0 Gemini 3.1 Pro
  11. 97.5 GPT-5.6 Sol
  12. 97.5 Claude Opus 5
  13. 97.5 Claude Fable 5.1
  14. 97.5 GPT-6.2 Sol
  15. 96.5 GPT-5.5 Pro
  16. 96.5 GPT-5.6 Terra
  17. 96.5 Claude Sonnet 5.5
  18. 96.5 Inkling 2 Small
  19. 96.0 Gemini 3 Deep Think
  20. 95.5 Gemini 3.7 Flash
  21. 95.5 GPT-6 Sol
  22. 95.0 GPT-5.5
  23. 94.5 GPT-5.4 Pro
  24. 94.5 Kimi K3
  25. 94.5 DeepSeek V4.1 Flash
  26. 94.5 DeepSeek V4.1 Flash 1017
  27. 94.0 Claude Opus 4.6
  28. 93.7 GPT-5.4
  29. 93.5 Claude Opus 4.7
  30. 92.5 Gemini 3.5 Flash
  31. 92.5 Claude Opus 4.8
  32. 91.2 Gemini 3.6 Flash
  33. 91.0 GLM-5.3-Flash
  34. 90.7 GPT-5.6 Luna
  35. 90.5 GPT-5.2 Pro
  36. 90.5 DeepSeek V4 Pro 0813
  37. 90.2 Grok 4.7
  38. 89.5 Grok 4.20
  39. 89.0 DeepSeek V4 Flash 0731
  40. 89.0 Claude Haiku 5.5
  41. 89.0 Muse Spark Open
  42. 88.5 Mercury 3
  43. 87.5 Grok 4.6
  44. 87.5 Qwen 3.8 27B
  45. 87.2 Grok 4.5
  46. 86.7 GPT-6 Luna
  47. 86.5 Claude Sonnet 4.6
  48. 86.2 GPT-5.2
  49. 84.7 Gemini 3 Flash
  50. 84.0 Inkling-Small
  51. 80.0 Claude Opus 4.5
  52. 79.5 Inkling
  53. 77.0 GLM-5.2
  54. 75.0 Gemini 3 Pro
  55. 74.0 Beam 1.0
  56. 72.8 GPT-5.1
  57. 70.2 GPT-5 Pro
  58. 66.7 Grok 4
  59. 65.7 GPT-5
  60. 65.3 Kimi K2.5
  61. 63.7 Claude Sonnet 4.5
  62. 63.7 MiniMax-M2.5
  63. 63.7 GPT-5.4 Mini
  64. 60.8 o3
  65. 59.3 o3-pro
  66. 58.7 o4-mini
  67. 57.0 DeepSeek-V3.2
  68. 54.3 GPT-5 mini
  69. 53.5 Gemini 3.5 Flash-Lite
  70. 51.5 GPT-5.4 Nano
  71. 48.5 Grok 4 Fast
  72. 47.7 Claude Haiku 4.5
  73. 44.7 GLM-5
  74. 41.0 Gemini 2.5 Pro
  75. 40.0 Claude Sonnet 4
  76. 40.0 tiny-recursion-model
  77. 35.7 Claude Opus 4
  78. 34.5 o3-mini
  79. 33.3 Gemini 2.5 Flash
  80. 28.6 Claude 3.7 Sonnet
  81. 27.3 codex-mini-2025-05-16
  82. 22.7 Nemotron 3.5 Super VL
  83. 21.2 DeepSeek-R1
  84. 20.7 GPT-5 nano
  85. 16.5 Grok-3 mini
  86. 15.8 DeepSeek-R1
  87. 14.0 o1-mini
  88. 11.0 Qwen3-235B-A22B-Instruct
  89. 10.3 GPT-4.5
  90. 6.12 Magistral Medium 1.0
  91. 5.50 Grok 3
  92. 5.50 GPT-4.1
  93. 5.00 Magistral Small 1.0
  94. 4.50 GPT-4o
  95. 4.38 Llama 4 Maverick
  96. 3.50 GPT-4.1 mini
  97. 0.50 Llama 4 Scout
  98. 0.0 GPT-4.1 nano
  1. 1 Claude Fable 5 max 98.5%
  2. 2 Gemini 3.8 Flash high 98.5%
  3. 3 GPT-6 Astra xhigh 98.5%
  4. 4 Claude Opus 5.5 high 98.5%
  5. 5 GPT-6.1 Sol xhigh 98.5%
  6. 6 Gemini 4 Argon high 98.5%
  7. 7 Gemini 4 Carbon medium 98.5%
  8. 8 Claude Fable 5.2 xhigh 98.5%
  9. 9 GPT-6.2 Astra max 98.5%
  10. 10 Gemini 3.1 Pro 98%
  11. 11 GPT-5.6 Sol xhigh 97.5%
  12. 12 Claude Opus 5 high 97.5%
  13. 13 Claude Fable 5.1 max 97.5%
  14. 14 GPT-6.2 Sol max 97.5%
  15. 15 GPT-5.5 Pro high 96.5%
  16. 16 GPT-5.6 Terra max 96.5%
  17. 17 Claude Sonnet 5.5 max 96.5%
  18. 18 Inkling 2 Small xhigh 96.5%
  19. 19 Gemini 3 Deep Think 96%
  20. 20 Gemini 3.7 Flash high 95.5%
  21. 21 GPT-6 Sol max 95.5%
  22. 22 GPT-5.5 xhigh 95%
  23. 23 GPT-5.4 Pro xhigh 94.5%
  24. 24 Kimi K3 max 94.5%
  25. 25 DeepSeek V4.1 Flash max 94.5%
  26. 26 DeepSeek V4.1 Flash 1017 high 94.5%
  27. 27 Claude Opus 4.6 120k, high 94%
  28. 28 GPT-5.4 xhigh 93.7%
  29. 29 Claude Opus 4.7 high 93.5%
  30. 30 Gemini 3.5 Flash high 92.5%
  31. 31 Claude Opus 4.8 max 92.5%
  32. 32 Gemini 3.6 Flash high 91.2%
  33. 33 GLM-5.3-Flash max 91%
  34. 34 GPT-5.6 Luna max 90.7%
  35. 35 GPT-5.2 Pro xhigh 90.5%
  36. 36 DeepSeek V4 Pro 0813 low 90.5%
  37. 37 Grok 4.7 high 90.2%
  38. 38 Grok 4.20 reasoning 89.5%
  39. 39 DeepSeek V4 Flash 0731 max 89%
  40. 40 Claude Haiku 5.5 max 89%
  41. 41 Muse Spark Open xhigh 89%
  42. 42 Mercury 3 high 88.5%
  43. 43 Grok 4.6 medium 87.5%
  44. 44 Qwen 3.8 27B xhigh 87.5%
  45. 45 Grok 4.5 medium 87.2%
  46. 46 GPT-6 Luna max 86.7%
  47. 47 Claude Sonnet 4.6 high 86.5%
  48. 48 GPT-5.2 xhigh 86.2%
  49. 49 Gemini 3 Flash high 84.7%
  50. 50 Inkling-Small xhigh 84%
  51. 51 Claude Opus 4.5 thinking, 64k 80%
  52. 52 Inkling 79.5%
  53. 53 GLM-5.2 77%
  54. 54 Gemini 3 Pro 75%
  55. 55 Beam 1.0 xhigh 74%
  56. 56 GPT-5.1 thinking, high 72.8%
  57. 57 GPT-5 Pro 70.2%
  58. 58 Grok 4 thinking 66.7%
  59. 59 GPT-5 high 65.7%
  60. 60 Kimi K2.5 65.3%
  61. 61 Claude Sonnet 4.5 thinking 32k 63.7%
  62. 62 MiniMax-M2.5 63.7%
  63. 63 GPT-5.4 Mini xhigh 63.7%
  64. 64 o3 high 60.8%
  65. 65 o3-pro high 59.3%
  66. 66 o4-mini high 58.7%
  67. 67 DeepSeek-V3.2 57%
  68. 68 GPT-5 mini high 54.3%
  69. 69 Gemini 3.5 Flash-Lite high 53.5%
  70. 70 GPT-5.4 Nano xhigh 51.5%
  71. 71 Grok 4 Fast 48.5%
  72. 72 Claude Haiku 4.5 thinking 32k 47.7%
  73. 73 GLM-5 44.7%
  74. 74 Gemini 2.5 Pro Jun 2025, thinking 16k 41%
  75. 75 Claude Sonnet 4 thinking 16k 40%
  76. 76 tiny-recursion-model 40%
  77. 77 Claude Opus 4 thinking 16k 35.7%
  78. 78 o3-mini high 34.5%
  79. 79 Gemini 2.5 Flash May 2025 33.3%
  80. 80 Claude 3.7 Sonnet 16k 28.6%
  81. 81 codex-mini-2025-05-16 27.3%
  82. 82 Nemotron 3.5 Super VL high 22.7%
  83. 83 DeepSeek-R1 May 2025 21.2%
  84. 84 GPT-5 nano medium 20.7%
  85. 85 Grok-3 mini low 16.5%
  86. 86 DeepSeek-R1 15.8%
  87. 87 o1-mini 14%
  88. 88 Qwen3-235B-A22B-Instruct Jul 2025 11%
  89. 89 GPT-4.5 10.3%
  90. 90 Magistral Medium 1.0 thinking 6.12%
  91. 91 Grok 3 5.50%
  92. 92 GPT-4.1 5.50%
  93. 93 Magistral Small 1.0 5.00%
  94. 94 GPT-4o Nov 2024 4.50%
  95. 95 Llama 4 Maverick 4.38%
  96. 96 GPT-4.1 mini 3.50%
  97. 97 Llama 4 Scout 0.50%
  98. 98 GPT-4.1 nano 0%

Source: Official ARC-AGI leaderboard

109 models. Scroll sideways for the rest.

  1. 98.5 Claude Fable 5
  2. 98.5 Gemini 3.8 Flash
  3. 98.5 GPT-6 Astra
  4. 98.5 Claude Opus 5.5
  5. 98.5 GPT-6.1 Sol
  6. 98.5 Gemini 4 Argon
  7. 98.5 Gemini 4 Carbon
  8. 98.5 Claude Fable 5.2
  9. 98.5 GPT-6.2 Astra
  10. 98.0 Gemini 3.1 Pro
  11. 97.5 GPT-5.6 Sol
  12. 97.5 Claude Opus 5
  13. 97.5 Claude Fable 5.1
  14. 97.5 GPT-6.2 Sol
  15. 96.5 GPT-5.5 Pro
  16. 96.5 GPT-5.6 Terra
  17. 96.5 Claude Sonnet 5.5
  18. 96.5 Inkling 2 Small
  19. 96.0 Gemini 3 Deep Think
  20. 95.5 Gemini 3.7 Flash
  21. 95.5 GPT-6 Sol
  22. 95.0 GPT-5.5
  23. 94.5 GPT-5.4 Pro
  24. 94.5 Kimi K3
  25. 94.5 DeepSeek V4.1 Flash
  26. 94.5 DeepSeek V4.1 Flash 1017
  27. 94.0 Claude Opus 4.6
  28. 93.7 GPT-5.4
  29. 93.5 Claude Opus 4.7
  30. 92.5 Gemini 3.5 Flash
  31. 92.5 Claude Opus 4.8
  32. 91.2 Gemini 3.6 Flash
  33. 91.0 GLM-5.3-Flash
  34. 90.7 GPT-5.6 Luna
  35. 90.5 GPT-5.2 Pro
  36. 90.5 DeepSeek V4 Pro 0813
  37. 90.2 Grok 4.7
  38. 89.5 Grok 4.20
  39. 89.0 DeepSeek V4 Flash 0731
  40. 89.0 Claude Haiku 5.5
  41. 89.0 Muse Spark Open
  42. 88.5 Mercury 3
  43. 87.5 Grok 4.6
  44. 87.5 Qwen 3.8 27B
  45. 87.2 Grok 4.5
  46. 86.7 GPT-6 Luna
  47. 86.5 Claude Sonnet 4.6
  48. 86.2 GPT-5.2
  49. 85.4 Seed2.0 Pro
  50. 84.7 Gemini 3 Flash
  51. 84.5 Inkling 2 Mini
  52. 84.0 Inkling-Small
  53. 80.0 Claude Opus 4.5
  54. 79.5 Inkling
  55. 77.0 GLM-5.2
  56. 75.0 Gemini 3 Pro
  57. 74.0 Beam 1.0
  58. 72.8 GPT-5.1
  59. 70.2 GPT-5 Pro
  60. 67.9 Seed1.8
  61. 66.9 Muse Glimmer
  62. 66.7 Grok 4
  63. 65.7 GPT-5
  64. 65.3 Kimi K2.5
  65. 63.7 Claude Sonnet 4.5
  66. 63.7 MiniMax-M2.5
  67. 63.7 GPT-5.4 Mini
  68. 60.8 o3
  69. 59.3 o3-pro
  70. 58.7 o4-mini
  71. 57.0 DeepSeek-V3.2
  72. 55.9 Ring-1T
  73. 54.8 Step 3.5 Flash
  74. 54.3 GPT-5 mini
  75. 53.5 Gemini 3.5 Flash-Lite
  76. 52.9 Tencent HY 2.0 Think
  77. 51.5 GPT-5.4 Nano
  78. 50.9 Ling-2.6-1T
  79. 50.3 LongCat-Flash-Thinking
  80. 48.5 Grok 4 Fast
  81. 47.7 Claude Haiku 4.5
  82. 44.7 GLM-5
  83. 41.0 Gemini 2.5 Pro
  84. 40.0 Claude Sonnet 4
  85. 40.0 tiny-recursion-model
  86. 39.9 Seed-Thinking-v1.5
  87. 35.7 Claude Opus 4
  88. 34.5 o3-mini
  89. 33.3 Gemini 2.5 Flash
  90. 28.6 Claude 3.7 Sonnet
  91. 27.3 codex-mini-2025-05-16
  92. 22.7 Nemotron 3.5 Super VL
  93. 21.2 DeepSeek-R1
  94. 20.7 GPT-5 nano
  95. 16.5 Grok-3 mini
  96. 15.8 DeepSeek-R1
  97. 14.0 o1-mini
  98. 11.0 Qwen3-235B-A22B-Instruct
  99. 10.3 GPT-4.5
  100. 6.12 Magistral Medium 1.0
  101. 5.91 Magistral Medium 1.1
  102. 5.50 Grok 3
  103. 5.50 GPT-4.1
  104. 5.00 Magistral Small 1.0
  105. 4.50 GPT-4o
  106. 4.38 Llama 4 Maverick
  107. 3.50 GPT-4.1 mini
  108. 0.50 Llama 4 Scout
  109. 0.0 GPT-4.1 nano
  1. 1 Claude Fable 5 max 98.5%
  2. 2 Gemini 3.8 Flash high 98.5%
  3. 3 GPT-6 Astra xhigh 98.5%
  4. 4 Claude Opus 5.5 high 98.5%
  5. 5 GPT-6.1 Sol xhigh 98.5%
  6. 6 Gemini 4 Argon high 98.5%
  7. 7 Gemini 4 Carbon medium 98.5%
  8. 8 Claude Fable 5.2 xhigh 98.5%
  9. 9 GPT-6.2 Astra max 98.5%
  10. 10 Gemini 3.1 Pro 98%
  11. 11 GPT-5.6 Sol xhigh 97.5%
  12. 12 Claude Opus 5 high 97.5%
  13. 13 Claude Fable 5.1 max 97.5%
  14. 14 GPT-6.2 Sol max 97.5%
  15. 15 GPT-5.5 Pro high 96.5%
  16. 16 GPT-5.6 Terra max 96.5%
  17. 17 Claude Sonnet 5.5 max 96.5%
  18. 18 Inkling 2 Small xhigh 96.5%
  19. 19 Gemini 3 Deep Think 96%
  20. 20 Gemini 3.7 Flash high 95.5%
  21. 21 GPT-6 Sol max 95.5%
  22. 22 GPT-5.5 xhigh 95%
  23. 23 GPT-5.4 Pro xhigh 94.5%
  24. 24 Kimi K3 max 94.5%
  25. 25 DeepSeek V4.1 Flash max 94.5%
  26. 26 DeepSeek V4.1 Flash 1017 high 94.5%
  27. 27 Claude Opus 4.6 120k, high 94%
  28. 28 GPT-5.4 93.7%
  29. 29 Claude Opus 4.7 high 93.5%
  30. 30 Gemini 3.5 Flash high 92.5%
  31. 31 Claude Opus 4.8 max 92.5%
  32. 32 Gemini 3.6 Flash high 91.2%
  33. 33 GLM-5.3-Flash max 91%
  34. 34 GPT-5.6 Luna max 90.7%
  35. 35 GPT-5.2 Pro xhigh 90.5%
  36. 36 DeepSeek V4 Pro 0813 low 90.5%
  37. 37 Grok 4.7 high 90.2%
  38. 38 Grok 4.20 reasoning 89.5%
  39. 39 DeepSeek V4 Flash 0731 max 89%
  40. 40 Claude Haiku 5.5 max 89%
  41. 41 Muse Spark Open xhigh 89%
  42. 42 Mercury 3 high 88.5%
  43. 43 Grok 4.6 medium 87.5%
  44. 44 Qwen 3.8 27B xhigh 87.5%
  45. 45 Grok 4.5 medium 87.2%
  46. 46 GPT-6 Luna max 86.7%
  47. 47 Claude Sonnet 4.6 high 86.5%
  48. 48 GPT-5.2 86.2%
  49. 49 Seed2.0 Pro 85.4%
  50. 50 Gemini 3 Flash high 84.7%
  51. 51 Inkling 2 Mini 84.5%
  52. 52 Inkling-Small xhigh 84%
  53. 53 Claude Opus 4.5 thinking, 64k 80%
  54. 54 Inkling 79.5%
  55. 55 GLM-5.2 77%
  56. 56 Gemini 3 Pro 75%
  57. 57 Beam 1.0 xhigh 74%
  58. 58 GPT-5.1 thinking, high 72.8%
  59. 59 GPT-5 Pro 70.2%
  60. 60 Seed1.8 67.9%
  61. 61 Muse Glimmer 66.9%
  62. 62 Grok 4 thinking 66.7%
  63. 63 GPT-5 high 65.7%
  64. 64 Kimi K2.5 65.3%
  65. 65 Claude Sonnet 4.5 thinking 32k 63.7%
  66. 66 MiniMax-M2.5 63.7%
  67. 67 GPT-5.4 Mini xhigh 63.7%
  68. 68 o3 high 60.8%
  69. 69 o3-pro high 59.3%
  70. 70 o4-mini high 58.7%
  71. 71 DeepSeek-V3.2 57%
  72. 72 Ring-1T 55.9%
  73. 73 Step 3.5 Flash 54.8%
  74. 74 GPT-5 mini high 54.3%
  75. 75 Gemini 3.5 Flash-Lite high 53.5%
  76. 76 Tencent HY 2.0 Think 52.9%
  77. 77 GPT-5.4 Nano xhigh 51.5%
  78. 78 Ling-2.6-1T 50.9%
  79. 79 LongCat-Flash-Thinking 50.3%
  80. 80 Grok 4 Fast 48.5%
  81. 81 Claude Haiku 4.5 thinking 32k 47.7%
  82. 82 GLM-5 44.7%
  83. 83 Gemini 2.5 Pro Jun 2025, thinking 16k 41%
  84. 84 Claude Sonnet 4 thinking 16k 40%
  85. 85 tiny-recursion-model 40%
  86. 86 Seed-Thinking-v1.5 39.9%
  87. 87 Claude Opus 4 thinking 16k 35.7%
  88. 88 o3-mini high 34.5%
  89. 89 Gemini 2.5 Flash May 2025 33.3%
  90. 90 Claude 3.7 Sonnet 16k 28.6%
  91. 91 codex-mini-2025-05-16 27.3%
  92. 92 Nemotron 3.5 Super VL high 22.7%
  93. 93 DeepSeek-R1 May 2025 21.2%
  94. 94 GPT-5 nano medium 20.7%
  95. 95 Grok-3 mini low 16.5%
  96. 96 DeepSeek-R1 15.8%
  97. 97 o1-mini 14%
  98. 98 Qwen3-235B-A22B-Instruct Jul 2025 11%
  99. 99 GPT-4.5 10.3%
  100. 100 Magistral Medium 1.0 thinking 6.12%
  101. 101 Magistral Medium 1.1 5.91%
  102. 102 Grok 3 5.50%
  103. 103 GPT-4.1 5.50%
  104. 104 Magistral Small 1.0 5.00%
  105. 105 GPT-4o Nov 2024 4.50%
  106. 106 Llama 4 Maverick 4.38%
  107. 107 GPT-4.1 mini 3.50%
  108. 108 Llama 4 Scout 0.50%
  109. 109 GPT-4.1 nano 0%

Sources: Official ARC-AGI leaderboard, labs’ model cards and another board, via Epoch AI (CC BY 4.0)

#EntryScoreSettingReleasedRun onSource
1 Claude Fable 5 98.5% max Jun 2026 Jun 2026 Official leaderboard
2 Gemini 3.8 Flash 98.5% high Sep 2026 Sep 2026 Official leaderboard
3 GPT-6 Astra 98.5% xhigh Sep 2026 Sep 2026 Official leaderboard
4 Claude Opus 5.5 98.5% high Sep 2026 Sep 2026 Official leaderboard
5 GPT-6.1 Sol 98.5% xhigh Sep 2026 Sep 2026 Official leaderboard
6 Gemini 4 Argon 98.5% high Sep 2026 Oct 2026 Official leaderboard
7 Gemini 4 Carbon 98.5% medium Oct 2026 Nov 2026 Official leaderboard
8 Claude Fable 5.2 98.5% xhigh Nov 2026 Nov 2026 Official leaderboard
9 GPT-6.2 Astra 98.5% max Nov 2026 Nov 2026 Official leaderboard
10 Gemini 3.1 Pro 98% Feb 2026 Feb 2026 Official leaderboard
11 GPT-5.6 Sol 97.5% xhigh Jul 2026 Jul 2026 Official leaderboard
12 Claude Opus 5 97.5% high Jul 2026 Jul 2026 Official leaderboard
13 Claude Fable 5.1 97.5% max Sep 2026 Sep 2026 Official leaderboard
14 GPT-6.2 Sol 97.5% max Nov 2026 Nov 2026 Official leaderboard
15 GPT-5.5 Pro 96.5% high Apr 2026 Apr 2026 Official leaderboard
16 GPT-5.6 Terra 96.5% max Jul 2026 Jul 2026 Official leaderboard
17 Claude Sonnet 5.5 96.5% max Sep 2026 Sep 2026 Official leaderboard
18 Inkling 2 Small 96.5% xhigh Nov 2026 Nov 2026 Official leaderboard
19 Gemini 3 Deep Think 96% Feb 2026 Feb 2026 Official leaderboard
20 Gemini 3.7 Flash 95.5% high Aug 2026 Aug 2026 Official leaderboard
21 GPT-6 Sol 95.5% max Sep 2026 Sep 2026 Official leaderboard
22 GPT-5.5 95% xhigh Apr 2026 Apr 2026 Official leaderboard
23 GPT-5.2 (Refinement) 94.5% Feb 2026 Official leaderboard
24 GPT-5.4 Pro 94.5% xhigh Mar 2026 Mar 2026 Official leaderboard
25 Kimi K3 94.5% max Jul 2026 Jul 2026 Official leaderboard
26 DeepSeek V4.1 Flash 94.5% max Sep 2026 Sep 2026 Official leaderboard
27 DeepSeek V4.1 Flash 1017 94.5% high Oct 2026 Oct 2026 Official leaderboard
28 Claude Opus 4.6 94% 120k, high Feb 2026 Feb 2026 Official leaderboard
29 GPT-5.4 93.7% xhigh Mar 2026 Mar 2026 Official leaderboard
30 Claude Opus 4.7 93.5% high Apr 2026 Apr 2026 Official leaderboard
31 Gemini 3.5 Flash 92.5% high May 2026 May 2026 Official leaderboard
32 Claude Opus 4.8 92.5% max May 2026 Jun 2026 Official leaderboard
33 Gemini 3.6 Flash 91.2% high Jul 2026 Jul 2026 Official leaderboard
34 GLM-5.3-Flash 91% max Aug 2026 Aug 2026 Official leaderboard
35 GPT-5.6 Luna 90.7% max Jul 2026 Jul 2026 Official leaderboard
36 GPT-5.2 Pro 90.5% xhigh Dec 2025 Dec 2025 Official leaderboard
37 DeepSeek V4 Pro 0813 90.5% low Aug 2026 Aug 2026 Official leaderboard
38 Grok 4.7 90.2% high Sep 2026 Sep 2026 Official leaderboard
39 Grok 4.20 89.5% reasoning Feb 2026 Mar 2026 Official leaderboard
40 DeepSeek V4 Flash 0731 89% max Jul 2026 Jul 2026 Official leaderboard
41 Claude Haiku 5.5 89% max Oct 2026 Oct 2026 Official leaderboard
42 Muse Spark Open 89% xhigh Oct 2026 Oct 2026 Official leaderboard
43 Mercury 3 88.5% high Oct 2026 Oct 2026 Official leaderboard
44 Grok 4.6 87.5% medium Aug 2026 Aug 2026 Official leaderboard
45 Qwen 3.8 27B 87.5% xhigh Aug 2026 Aug 2026 Official leaderboard
46 Grok 4.5 87.2% medium Jul 2026 Jul 2026 Official leaderboard
47 GPT-6 Luna 86.7% max Sep 2026 Sep 2026 Official leaderboard
48 Claude Sonnet 4.6 86.5% high Feb 2026 Feb 2026 Official leaderboard
49 GPT-5.2 86.2% xhigh Dec 2025 Dec 2025 Official leaderboard
50 Gemini 3 Flash 84.7% high Dec 2025 Dec 2025 Official leaderboard
51 Inkling-Small 84% xhigh Jul 2026 Jul 2026 Official leaderboard
52 Claude Opus 4.5 80% thinking, 64k Nov 2025 Nov 2025 Official leaderboard
53 Grok 4 (Refinement) 79.6% Aug 2025 Official leaderboard
54 Inkling 79.5% Jul 2026 Jul 2026 Official leaderboard
55 GLM-5.2 77% Jun 2026 Jun 2026 Official leaderboard
56 Gemini 3 Pro 75% Nov 2025 Nov 2025 Official leaderboard
57 Beam 1.0 74% xhigh Oct 2026 Oct 2026 Official leaderboard
58 GPT-5.1 72.8% thinking, high Nov 2025 Nov 2025 Official leaderboard
59 GPT-5 Pro 70.2% Oct 2025 Oct 2025 Official leaderboard
60 Grok 4 66.7% thinking Jul 2025 Jul 2025 Official leaderboard
61 GPT-5 65.7% high Aug 2025 Aug 2025 Official leaderboard
62 Kimi K2.5 65.3% Jan 2026 Jan 2026 Official leaderboard
63 Claude Sonnet 4.5 63.7% thinking 32k Sep 2025 Sep 2025 Official leaderboard
64 MiniMax-M2.5 63.7% Feb 2026 Feb 2026 Official leaderboard
65 GPT-5.4 Mini 63.7% xhigh Mar 2026 Mar 2026 Official leaderboard
66 o3 60.8% high Apr 2025 Apr 2025 Official leaderboard
67 o3-pro 59.3% high Jun 2025 Jun 2025 Official leaderboard
68 o4-mini 58.7% high Apr 2025 Apr 2025 Official leaderboard
69 DeepSeek-V3.2 57% Dec 2025 Dec 2025 Official leaderboard
70 GPT-5 mini 54.3% high Aug 2025 Aug 2025 Official leaderboard
71 Gemini 3.5 Flash-Lite 53.5% high Jul 2026 Jul 2026 Official leaderboard
72 GPT-5.4 Nano 51.5% xhigh Mar 2026 Mar 2026 Official leaderboard
73 Grok 4 Fast 48.5% Sep 2025 Oct 2025 Official leaderboard
74 Claude Haiku 4.5 47.7% thinking 32k Oct 2025 Oct 2025 Official leaderboard
75 GLM-5 44.7% Feb 2026 Feb 2026 Official leaderboard
76 Gemini 2.5 Pro Jun 2025 41% thinking 16k Jun 2025 Jun 2025 Official leaderboard
77 Claude Sonnet 4 40% thinking 16k May 2025 May 2025 Official leaderboard
78 tiny-recursion-model 40% Oct 2025 Oct 2025 Official leaderboard
79 Claude Opus 4 35.7% thinking 16k May 2025 May 2025 Official leaderboard
80 o3-mini 34.5% high Jan 2025 Jan 2025 Official leaderboard
81 Gemini 2.5 Flash May 2025 33.3% May 2025 May 2025 Official leaderboard
82 Claude 3.7 Sonnet 28.6% 16k Feb 2025 Feb 2025 Official leaderboard
83 codex-mini-2025-05-16 27.3% May 2025 May 2025 Official leaderboard
84 Nemotron 3.5 Super VL 22.7% high Oct 2026 Oct 2026 Official leaderboard
85 DeepSeek-R1 May 2025 21.2% May 2025 May 2025 Official leaderboard
86 GPT-5 nano 20.7% medium Aug 2025 Aug 2025 Official leaderboard
87 Grok-3 mini 16.5% low Apr 2025 Feb 2025 Official leaderboard
88 DeepSeek-R1 15.8% Jan 2025 Jan 2025 Official leaderboard
89 o1-mini 14% Sep 2024 Sep 2024 Official leaderboard
90 Qwen3-235B-A22B-Instruct Jul 2025 11% Jul 2025 Jul 2025 Official leaderboard
91 Mercury 3 Preview (Instant) 10.7% Oct 2026 Official leaderboard
92 GPT-4.5 10.3% Feb 2025 Feb 2025 Official leaderboard
93 Magistral Medium 1.0 6.12% thinking Jun 2025 Jun 2025 Official leaderboard
94 Grok 3 5.50% Apr 2025 Feb 2025 Official leaderboard
95 GPT-4.1 5.50% Apr 2025 Apr 2025 Official leaderboard
96 Magistral Small 1.0 5.00% Jun 2025 Jun 2025 Official leaderboard
97 GPT-4o Nov 2024 4.50% Nov 2024 Nov 2024 Official leaderboard
98 Llama 4 Maverick 4.38% Apr 2025 Apr 2025 Official leaderboard
99 GPT-4.1 mini 3.50% Apr 2025 Apr 2025 Official leaderboard
100 Llama 4 Scout 0.50% Apr 2025 Apr 2025 Official leaderboard
101 GPT-4.1 nano 0% Apr 2025 Apr 2025 Official leaderboard
#ModelScoreSettingReleasedRun onSource
1 Claude Fable 5 98.5% max Jun 2026 Jun 2026 Official leaderboard
2 Gemini 3.8 Flash 98.5% high Sep 2026 Sep 2026 Official leaderboard
3 GPT-6 Astra 98.5% xhigh Sep 2026 Sep 2026 Official leaderboard
4 Claude Opus 5.5 98.5% high Sep 2026 Sep 2026 Official leaderboard
5 GPT-6.1 Sol 98.5% xhigh Sep 2026 Sep 2026 Official leaderboard
6 Gemini 4 Argon 98.5% high Sep 2026 Oct 2026 Official leaderboard
7 Gemini 4 Carbon 98.5% medium Oct 2026 Nov 2026 Official leaderboard
8 Claude Fable 5.2 98.5% xhigh Nov 2026 Nov 2026 Official leaderboard
9 GPT-6.2 Astra 98.5% max Nov 2026 Nov 2026 Official leaderboard
10 Gemini 3.1 Pro 98% Feb 2026 Feb 2026 Official leaderboard
11 GPT-5.6 Sol 97.5% xhigh Jul 2026 Jul 2026 Official leaderboard
12 Claude Opus 5 97.5% high Jul 2026 Jul 2026 Official leaderboard
13 Claude Fable 5.1 97.5% max Sep 2026 Sep 2026 Official leaderboard
14 GPT-6.2 Sol 97.5% max Nov 2026 Nov 2026 Official leaderboard
15 GPT-5.5 Pro 96.5% high Apr 2026 Apr 2026 Official leaderboard
16 GPT-5.6 Terra 96.5% max Jul 2026 Jul 2026 Official leaderboard
17 Claude Sonnet 5.5 96.5% max Sep 2026 Sep 2026 Official leaderboard
18 Inkling 2 Small 96.5% xhigh Nov 2026 Nov 2026 Official leaderboard
19 Gemini 3 Deep Think 96% Feb 2026 Feb 2026 Official leaderboard
20 Gemini 3.7 Flash 95.5% high Aug 2026 Aug 2026 Official leaderboard
21 GPT-6 Sol 95.5% max Sep 2026 Sep 2026 Official leaderboard
22 GPT-5.5 95% xhigh Apr 2026 Apr 2026 Official leaderboard
23 GPT-5.4 Pro 94.5% xhigh Mar 2026 Mar 2026 Official leaderboard
24 Kimi K3 94.5% max Jul 2026 Jul 2026 Official leaderboard
25 DeepSeek V4.1 Flash 94.5% max Sep 2026 Sep 2026 Official leaderboard
26 DeepSeek V4.1 Flash 1017 94.5% high Oct 2026 Oct 2026 Official leaderboard
27 Claude Opus 4.6 94% 120k, high Feb 2026 Feb 2026 Official leaderboard
28 GPT-5.4 93.7% xhigh Mar 2026 Mar 2026 Official leaderboard
29 Claude Opus 4.7 93.5% high Apr 2026 Apr 2026 Official leaderboard
30 Gemini 3.5 Flash 92.5% high May 2026 May 2026 Official leaderboard
31 Claude Opus 4.8 92.5% max May 2026 Jun 2026 Official leaderboard
32 Gemini 3.6 Flash 91.2% high Jul 2026 Jul 2026 Official leaderboard
33 GLM-5.3-Flash 91% max Aug 2026 Aug 2026 Official leaderboard
34 GPT-5.6 Luna 90.7% max Jul 2026 Jul 2026 Official leaderboard
35 GPT-5.2 Pro 90.5% xhigh Dec 2025 Dec 2025 Official leaderboard
36 DeepSeek V4 Pro 0813 90.5% low Aug 2026 Aug 2026 Official leaderboard
37 Grok 4.7 90.2% high Sep 2026 Sep 2026 Official leaderboard
38 Grok 4.20 89.5% reasoning Feb 2026 Mar 2026 Official leaderboard
39 DeepSeek V4 Flash 0731 89% max Jul 2026 Jul 2026 Official leaderboard
40 Claude Haiku 5.5 89% max Oct 2026 Oct 2026 Official leaderboard
41 Muse Spark Open 89% xhigh Oct 2026 Oct 2026 Official leaderboard
42 Mercury 3 88.5% high Oct 2026 Oct 2026 Official leaderboard
43 Grok 4.6 87.5% medium Aug 2026 Aug 2026 Official leaderboard
44 Qwen 3.8 27B 87.5% xhigh Aug 2026 Aug 2026 Official leaderboard
45 Grok 4.5 87.2% medium Jul 2026 Jul 2026 Official leaderboard
46 GPT-6 Luna 86.7% max Sep 2026 Sep 2026 Official leaderboard
47 Claude Sonnet 4.6 86.5% high Feb 2026 Feb 2026 Official leaderboard
48 GPT-5.2 86.2% xhigh Dec 2025 Dec 2025 Official leaderboard
49 Gemini 3 Flash 84.7% high Dec 2025 Dec 2025 Official leaderboard
50 Inkling-Small 84% xhigh Jul 2026 Jul 2026 Official leaderboard
51 Claude Opus 4.5 80% thinking, 64k Nov 2025 Nov 2025 Official leaderboard
52 Inkling 79.5% Jul 2026 Jul 2026 Official leaderboard
53 GLM-5.2 77% Jun 2026 Jun 2026 Official leaderboard
54 Gemini 3 Pro 75% Nov 2025 Nov 2025 Official leaderboard
55 Beam 1.0 74% xhigh Oct 2026 Oct 2026 Official leaderboard
56 GPT-5.1 72.8% thinking, high Nov 2025 Nov 2025 Official leaderboard
57 GPT-5 Pro 70.2% Oct 2025 Oct 2025 Official leaderboard
58 Grok 4 66.7% thinking Jul 2025 Jul 2025 Official leaderboard
59 GPT-5 65.7% high Aug 2025 Aug 2025 Official leaderboard
60 Kimi K2.5 65.3% Jan 2026 Jan 2026 Official leaderboard
61 Claude Sonnet 4.5 63.7% thinking 32k Sep 2025 Sep 2025 Official leaderboard
62 MiniMax-M2.5 63.7% Feb 2026 Feb 2026 Official leaderboard
63 GPT-5.4 Mini 63.7% xhigh Mar 2026 Mar 2026 Official leaderboard
64 o3 60.8% high Apr 2025 Apr 2025 Official leaderboard
65 o3-pro 59.3% high Jun 2025 Jun 2025 Official leaderboard
66 o4-mini 58.7% high Apr 2025 Apr 2025 Official leaderboard
67 DeepSeek-V3.2 57% Dec 2025 Dec 2025 Official leaderboard
68 GPT-5 mini 54.3% high Aug 2025 Aug 2025 Official leaderboard
69 Gemini 3.5 Flash-Lite 53.5% high Jul 2026 Jul 2026 Official leaderboard
70 GPT-5.4 Nano 51.5% xhigh Mar 2026 Mar 2026 Official leaderboard
71 Grok 4 Fast 48.5% Sep 2025 Oct 2025 Official leaderboard
72 Claude Haiku 4.5 47.7% thinking 32k Oct 2025 Oct 2025 Official leaderboard
73 GLM-5 44.7% Feb 2026 Feb 2026 Official leaderboard
74 Gemini 2.5 Pro Jun 2025 41% thinking 16k Jun 2025 Jun 2025 Official leaderboard
75 Claude Sonnet 4 40% thinking 16k May 2025 May 2025 Official leaderboard
76 tiny-recursion-model 40% Oct 2025 Oct 2025 Official leaderboard
77 Claude Opus 4 35.7% thinking 16k May 2025 May 2025 Official leaderboard
78 o3-mini 34.5% high Jan 2025 Jan 2025 Official leaderboard
79 Gemini 2.5 Flash May 2025 33.3% May 2025 May 2025 Official leaderboard
80 Claude 3.7 Sonnet 28.6% 16k Feb 2025 Feb 2025 Official leaderboard
81 codex-mini-2025-05-16 27.3% May 2025 May 2025 Official leaderboard
82 Nemotron 3.5 Super VL 22.7% high Oct 2026 Oct 2026 Official leaderboard
83 DeepSeek-R1 May 2025 21.2% May 2025 May 2025 Official leaderboard
84 GPT-5 nano 20.7% medium Aug 2025 Aug 2025 Official leaderboard
85 Grok-3 mini 16.5% low Apr 2025 Feb 2025 Official leaderboard
86 DeepSeek-R1 15.8% Jan 2025 Jan 2025 Official leaderboard
87 o1-mini 14% Sep 2024 Sep 2024 Official leaderboard
88 Qwen3-235B-A22B-Instruct Jul 2025 11% Jul 2025 Jul 2025 Official leaderboard
89 GPT-4.5 10.3% Feb 2025 Feb 2025 Official leaderboard
90 Magistral Medium 1.0 6.12% thinking Jun 2025 Jun 2025 Official leaderboard
91 Grok 3 5.50% Apr 2025 Feb 2025 Official leaderboard
92 GPT-4.1 5.50% Apr 2025 Apr 2025 Official leaderboard
93 Magistral Small 1.0 5.00% Jun 2025 Jun 2025 Official leaderboard
94 GPT-4o Nov 2024 4.50% Nov 2024 Nov 2024 Official leaderboard
95 Llama 4 Maverick 4.38% Apr 2025 Apr 2025 Official leaderboard
96 GPT-4.1 mini 3.50% Apr 2025 Apr 2025 Official leaderboard
97 Llama 4 Scout 0.50% Apr 2025 Apr 2025 Official leaderboard
98 GPT-4.1 nano 0% Apr 2025 Apr 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Fable 5 98.5% max Jun 2026 Official leaderboard
2 Gemini 3.8 Flash 98.5% high Sep 2026 Official leaderboard
3 GPT-6 Astra 98.5% xhigh Sep 2026 Official leaderboard
4 Claude Opus 5.5 98.5% high Sep 2026 Official leaderboard
5 GPT-6.1 Sol 98.5% xhigh Sep 2026 Official leaderboard
6 Gemini 4 Argon 98.5% high Sep 2026 Official leaderboard
7 Gemini 4 Carbon 98.5% medium Oct 2026 Official leaderboard
8 Claude Fable 5.2 98.5% xhigh Nov 2026 Official leaderboard
9 GPT-6.2 Astra 98.5% max Nov 2026 Official leaderboard
10 Gemini 3.1 Pro 98% Feb 2026 Official leaderboard
11 GPT-5.6 Sol 97.5% xhigh Jul 2026 Official leaderboard
12 Claude Opus 5 97.5% high Jul 2026 Official leaderboard
13 Claude Fable 5.1 97.5% max Sep 2026 Official leaderboard
14 GPT-6.2 Sol 97.5% max Nov 2026 Official leaderboard
15 GPT-5.5 Pro 96.5% high Apr 2026 Official leaderboard
16 GPT-5.6 Terra 96.5% max Jul 2026 Official leaderboard
17 Claude Sonnet 5.5 96.5% max Sep 2026 Official leaderboard
18 Inkling 2 Small 96.5% xhigh Nov 2026 Official leaderboard
19 Gemini 3 Deep Think 96% Feb 2026 Official leaderboard
20 Gemini 3.7 Flash 95.5% high Aug 2026 Official leaderboard
21 GPT-6 Sol 95.5% max Sep 2026 Official leaderboard
22 GPT-5.5 95% xhigh Apr 2026 Official leaderboard
23 GPT-5.4 Pro 94.5% xhigh Mar 2026 Official leaderboard
24 Kimi K3 94.5% max Jul 2026 Official leaderboard
25 DeepSeek V4.1 Flash 94.5% max Sep 2026 Official leaderboard
26 DeepSeek V4.1 Flash 1017 94.5% high Oct 2026 Official leaderboard
27 Claude Opus 4.6 94% 120k, high Feb 2026 Official leaderboard
28 GPT-5.4 93.7% Mar 2026 Lab model card
29 Claude Opus 4.7 93.5% high Apr 2026 Official leaderboard
30 Gemini 3.5 Flash 92.5% high May 2026 Official leaderboard
31 Claude Opus 4.8 92.5% max May 2026 Official leaderboard
32 Gemini 3.6 Flash 91.2% high Jul 2026 Official leaderboard
33 GLM-5.3-Flash 91% max Aug 2026 Official leaderboard
34 GPT-5.6 Luna 90.7% max Jul 2026 Official leaderboard
35 GPT-5.2 Pro 90.5% xhigh Dec 2025 Official leaderboard
36 DeepSeek V4 Pro 0813 90.5% low Aug 2026 Official leaderboard
37 Grok 4.7 90.2% high Sep 2026 Official leaderboard
38 Grok 4.20 89.5% reasoning Feb 2026 Official leaderboard
39 DeepSeek V4 Flash 0731 89% max Jul 2026 Official leaderboard
40 Claude Haiku 5.5 89% max Oct 2026 Official leaderboard
41 Muse Spark Open 89% xhigh Oct 2026 Official leaderboard
42 Mercury 3 88.5% high Oct 2026 Official leaderboard
43 Grok 4.6 87.5% medium Aug 2026 Official leaderboard
44 Qwen 3.8 27B 87.5% xhigh Aug 2026 Official leaderboard
45 Grok 4.5 87.2% medium Jul 2026 Official leaderboard
46 GPT-6 Luna 86.7% max Sep 2026 Official leaderboard
47 Claude Sonnet 4.6 86.5% high Feb 2026 Official leaderboard
48 GPT-5.2 86.2% Dec 2025 Lab model card
49 Seed2.0 Pro 85.4% Feb 2026 Lab model card
50 Gemini 3 Flash 84.7% high Dec 2025 Official leaderboard
51 Inkling 2 Mini 84.5% Nov 2026 Lab model card
52 Inkling-Small 84% xhigh Jul 2026 Official leaderboard
53 Claude Opus 4.5 80% thinking, 64k Nov 2025 Official leaderboard
54 Inkling 79.5% Jul 2026 Official leaderboard
55 GLM-5.2 77% Jun 2026 Official leaderboard
56 Gemini 3 Pro 75% Nov 2025 Official leaderboard
57 Beam 1.0 74% xhigh Oct 2026 Official leaderboard
58 GPT-5.1 72.8% thinking, high Nov 2025 Official leaderboard
59 GPT-5 Pro 70.2% Oct 2025 Official leaderboard
60 Seed1.8 67.9% Dec 2025 Lab model card
61 Muse Glimmer 66.9% Oct 2026 Lab model card
62 Grok 4 66.7% thinking Jul 2025 Official leaderboard
63 GPT-5 65.7% high Aug 2025 Official leaderboard
64 Kimi K2.5 65.3% Jan 2026 Official leaderboard
65 Claude Sonnet 4.5 63.7% thinking 32k Sep 2025 Official leaderboard
66 MiniMax-M2.5 63.7% Feb 2026 Official leaderboard
67 GPT-5.4 Mini 63.7% xhigh Mar 2026 Official leaderboard
68 o3 60.8% high Apr 2025 Official leaderboard
69 o3-pro 59.3% high Jun 2025 Official leaderboard
70 o4-mini 58.7% high Apr 2025 Official leaderboard
71 DeepSeek-V3.2 57% Dec 2025 Official leaderboard
72 Ring-1T 55.9% Oct 2025 Lab model card
73 Step 3.5 Flash 54.8% Feb 2026 Lab model card
74 GPT-5 mini 54.3% high Aug 2025 Official leaderboard
75 Gemini 3.5 Flash-Lite 53.5% high Jul 2026 Official leaderboard
76 Tencent HY 2.0 Think 52.9% Dec 2025 Lab model card
77 GPT-5.4 Nano 51.5% xhigh Mar 2026 Official leaderboard
78 Ling-2.6-1T 50.9% Jun 2026 Lab model card
79 LongCat-Flash-Thinking 50.3% Sep 2025 Lab model card
80 Grok 4 Fast 48.5% Sep 2025 Official leaderboard
81 Claude Haiku 4.5 47.7% thinking 32k Oct 2025 Official leaderboard
82 GLM-5 44.7% Feb 2026 Official leaderboard
83 Gemini 2.5 Pro Jun 2025 41% thinking 16k Jun 2025 Official leaderboard
84 Claude Sonnet 4 40% thinking 16k May 2025 Official leaderboard
85 tiny-recursion-model 40% Oct 2025 Official leaderboard
86 Seed-Thinking-v1.5 39.9% Apr 2025 Lab model card
87 Claude Opus 4 35.7% thinking 16k May 2025 Official leaderboard
88 o3-mini 34.5% high Jan 2025 Official leaderboard
89 Gemini 2.5 Flash May 2025 33.3% May 2025 Official leaderboard
90 Claude 3.7 Sonnet 28.6% 16k Feb 2025 Official leaderboard
91 codex-mini-2025-05-16 27.3% May 2025 Official leaderboard
92 Nemotron 3.5 Super VL 22.7% high Oct 2026 Official leaderboard
93 DeepSeek-R1 May 2025 21.2% May 2025 Official leaderboard
94 GPT-5 nano 20.7% medium Aug 2025 Official leaderboard
95 Grok-3 mini 16.5% low Apr 2025 Official leaderboard
96 DeepSeek-R1 15.8% Jan 2025 Official leaderboard
97 o1-mini 14% Sep 2024 Official leaderboard
98 Qwen3-235B-A22B-Instruct Jul 2025 11% Jul 2025 Official leaderboard
99 GPT-4.5 10.3% Feb 2025 Official leaderboard
100 Magistral Medium 1.0 6.12% thinking Jun 2025 Official leaderboard
101 Magistral Medium 1.1 5.91% Jun 2025 Collected by Epoch AI, source not named
102 Grok 3 5.50% Apr 2025 Official leaderboard
103 GPT-4.1 5.50% Apr 2025 Official leaderboard
104 Magistral Small 1.0 5.00% Jun 2025 Official leaderboard
105 GPT-4o Nov 2024 4.50% Nov 2024 Official leaderboard
106 Llama 4 Maverick 4.38% Apr 2025 Official leaderboard
107 GPT-4.1 mini 3.50% Apr 2025 Official leaderboard
108 Llama 4 Scout 0.50% Apr 2025 Official leaderboard
109 GPT-4.1 nano 0% Apr 2025 Official leaderboard

Labs report scores for 22 models in their cards. Epoch AI collected claims for 30 models from labs’ reports and papers. Switch the scores above to rank them together.

The best score, 98.5%, costs $0.06 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI DeepSeek Z.ai Anthropic Inception Labs Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale GPT-6.1 Sol, high GPT-6.1 Sol, low DeepSeek V4 Flash 0731, max Mercury 3, medium GPT-6 Luna, high GPT-6 Luna, medium Claude Haiku 5.5, low 0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale GPT-6.1 Sol, high GPT-6.1 Sol, medium

Source: Official ARC-AGI leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5 max 98.5% $2.11
Claude Fable 5 xhigh 98.5% $1.02
Gemini 3.8 Flash high 98.5% $0.21
GPT-6 Astra xhigh 98.5% $0.35
GPT-6 Astra high 98.5% $0.28
Claude Opus 5.5 high 98.5% $0.16
GPT-6.1 Sol xhigh 98.5% $0.07
GPT-6.1 Sol high 98.5% $0.06
Gemini 4 Argon high 98.5% $0.16
Gemini 4 Argon medium 98.5% $0.06
Gemini 4 Carbon medium 98.5% $0.16
Claude Fable 5.2 xhigh 98.5% $0.36
GPT-6.2 Astra max 98.5% $0.35
GPT-6.2 Astra xhigh 98.5% $0.22
GPT-6.2 Astra high 98.5% $0.20
GPT-6.2 Astra medium 98.5% $0.17
GPT-6.2 Astra low 98.5% $0.17
Gemini 3.1 Pro 98% $0.52
GPT-5.6 Sol xhigh 97.5% $0.40
Claude Opus 5 high 97.5% $0.47
Claude Opus 5 max 97.5% $0.70
Claude Fable 5.1 max 97.5% $1.40
Gemini 3.8 Flash medium 97.5% $0.11
GPT-6 Astra max 97.5% $0.43
GPT-6 Astra medium 97.5% $0.24
Claude Opus 5.5 max 97.5% $0.64
Claude Opus 5.5 xhigh 97.5% $0.22
Claude Opus 5.5 medium 97.5% $0.15
Gemini 4 Carbon high 97.5% $0.41
Claude Fable 5.2 max 97.5% $0.92
Claude Fable 5.2 high 97.5% $0.33
Claude Fable 5.2 medium 97.5% $0.26
GPT-6.2 Sol max 97.5% $0.11
GPT-6.2 Sol high 97.5% $0.06
GPT-5.6 Sol high 97% $0.30
GPT-6.2 Sol xhigh 97% $0.08
GPT-5.5 Pro high 96.5% $4.53
GPT-5.6 Sol max 96.5% $0.54
GPT-5.6 Terra max 96.5% $0.55
Claude Fable 5.1 xhigh 96.5% $0.96
GPT-6 Astra low 96.5% $0.23
Claude Sonnet 5.5 max 96.5% $0.50
Claude Sonnet 5.5 xhigh 96.5% $0.14
GPT-6.1 Sol max 96.5% $0.10
Inkling 2 Small xhigh 96.5% $0.08
Gemini 3 Deep Think 96% $7.17
Claude Fable 5.1 high 96% $0.52
Claude Fable 5 high 95.5% $0.77
Gemini 3.7 Flash high 95.5% $0.12
GPT-6 Sol max 95.5% $0.14
GPT-6.1 Sol medium 95.5% $0.05
GPT-5.5 xhigh 95% $0.73
GPT-5.5 Pro xhigh 95% $4.52
GPT-5.4 Pro xhigh 94.5% $7.21
GPT-5.5 high 94.5% $0.56
Kimi K3 max 94.5% $0.77
Claude Fable 5.1 medium 94.5% $0.44
DeepSeek V4.1 Flash max 94.5% $0.07
DeepSeek V4.1 Flash 1017 high 94.5% $0.06
Claude Fable 5.2 low 94.5% $0.23
GPT-6.2 Sol medium 94.5% $0.05
Claude Opus 4.6 120k, high 94% $1.40
GPT-5.6 Terra xhigh 94% $0.26
GPT-5.4 xhigh 93.7% $0.62
Claude Opus 4.7 high 93.5% $1.41
GPT-6.1 Sol low 93.5% $0.05
DeepSeek V4.1 Flash 1017 max 93.5% $0.06
Claude Opus 4.6 120k, max 93% $1.88
GPT-5.4 high 92.7% $0.37
GPT-6 Sol xhigh 92.7% $0.10
Gemini 3.5 Flash high 92.5% $0.43
Claude Opus 4.8 max 92.5% $2.33
Claude Fable 5 medium 92.5% $0.66
GPT-5.6 Sol medium 92.5% $0.22
GPT-5.5 medium 92.2% $0.39
Claude Opus 4.6 120k, medium 92% $0.96
Claude Opus 4.7 max 92% $2.58
Claude Opus 4.8 high 92% $1.04
GPT-5.6 Terra high 92% $0.19
Claude Sonnet 5.5 high 92% $0.11
Claude Opus 4.8 medium 91.5% $0.91
DeepSeek V4.1 Flash 1017 low 91.5% $0.05
Gemini 3.6 Flash high 91.2% $0.34
Gemini 3.7 Flash medium 91.2% $0.06
Claude Opus 4.7 low 91% $0.76
Claude Opus 4.7 medium 91% $1.04
GLM-5.3-Flash max 91% $0.04
GPT-6 Sol high 91% $0.08
GPT-5.6 Luna max 90.7% $0.07
GPT-5.2 Pro xhigh 90.5% $12
Claude Fable 5 low 90.5% $0.51
DeepSeek V4 Pro 0813 low 90.5% $0.18
Gemini 3.8 Flash low 90.5% $0.07
DeepSeek V4.1 Flash low 90.5% $0.05
Grok 4.7 high 90.2% $0.64
DeepSeek V4 Pro 0813 max 90% $0.30
Claude Fable 5.1 low 90% $0.37
Grok 4.7 medium 90% $0.52
Grok 4.20 reasoning 89.5% $0.35
Grok 4.7 xhigh 89.5% $0.78
DeepSeek V4 Flash 0731 max 89% $0.02
Claude Haiku 5.5 max 89% $0.05
Muse Spark Open xhigh 89% $0.29
DeepSeek V4.1 Flash high 88.5% $0.06
Claude Opus 5.5 low 88.5% $0.12
Muse Spark Open high 88.5% $0.23
Mercury 3 high 88.5% $0.03
Claude Opus 4.8 low 88% $0.67
GPT-5.6 Luna max 88% $0.32
GPT-5.6 Luna xhigh 87.7% $0.21
Gemini 3 Deep Think 87.5% $44
Grok 4.6 medium 87.5% $0.30
Qwen 3.8 27B xhigh 87.5% $0.22
Claude Haiku 5.5 xhigh 87.5% $0.03
Grok 4.5 medium 87.2% $0.31
DeepSeek V4 Pro 0813 high 87.2% $0.31
DeepSeek V4 Flash 0731 high 87% $0.02
Grok 4.6 xhigh 87% $0.35
Grok 4.6 high 87% $0.33
Kimi K3 high 86.7% $0.48
GPT-6 Luna max 86.7% $0.02
Claude Sonnet 4.6 high 86.5% $1.19
GPT-5.2 xhigh 86.2% $0.96
GPT-5.4 medium 86.2% $0.25
Claude Opus 4.6 120k, low 86% $0.65
Claude Sonnet 4.6 max 86% $1.45
GPT-6 Astra none 86% $0.21
GPT-5.2 Pro high 85.7% $5.87
Grok 4.5 high 85.7% $0.33
Gemini 4 Carbon low 85.5% $0.06
Gemini 3.7 Flash low 85.2% $0.04
GPT-5.6 Luna xhigh 85% $0.05
Gemini 3 Flash high 84.7% $0.17
Mercury 3 medium 84.7% $0.02
Inkling-Small xhigh 84% $0.12
DeepSeek V4 Flash 0731 low 84% $0.01
GPT-6.2 Sol low 84% $0.04
GPT-6 Sol medium 83.7% $0.07
Gemini 3.6 Flash medium 83.2% $0.25
Claude Haiku 5.5 high 81.5% $0.02
GPT-5.2 Pro medium 81.2% $3.98
Mercury 3 low 80.7% $0.02
Claude Opus 4.5 thinking, 64k 80% $1.47
Inkling 79.5% $0.30
Grok 4.5 low 79.2% $0.18
GPT-5.2 high 78.7% $0.52
GPT-5.6 Luna high 78% $0.03
Inkling-Small high 78% $0.07
Muse Spark Open medium 77.7% $0.12
GLM-5.2 77% $0.19
GPT-5.6 Terra medium 77% $0.13
GPT-5.6 Luna high 76.5% $0.13
Gemini 3.6 Flash low 76.5% $0.12
Gemini 4 Argon low 76.5% $0.04
GPT-5.5 low 76.2% $0.20
Claude Opus 4.5 thinking, 32k 75.8% $0.95
Gemini 3 Pro 75% $0.49
Grok 4.6 low 74.8% $0.10
GPT-5.6 Sol low 74.5% $0.17
Beam 1.0 xhigh 74% $0.31
GPT-6 Luna xhigh 73% $0.01
GPT-5.1 thinking, high 72.8% $0.67
GPT-5.2 medium 72.7% $0.34
GPT-6 Sol low 72.2% $0.05
Claude Opus 4.5 thinking, 16k 72% $0.62
GLM-5.3-Flash high 71.8% $0.03
GPT-6 Luna high 70.3% $0.009
GPT-5 Pro 70.2% $4.78
Claude Sonnet 5.5 medium 70% $0.07
Qwen 3.8 27B low 69.2% $0.12
Qwen 3.8 27B medium 68.7% $0.13
GPT-5.4 low 68.2% $0.15
Beam 1.0 high 67.7% $0.20
Inkling-Small medium 67% $0.03
Claude Sonnet 5.5 low 67% $0.06
Grok 4 thinking 66.7% $1.01
GPT-5 high 65.7% $0.51
Kimi K3 low 65.7% $0.18
Kimi K2.5 65.3% $0.15
Claude Sonnet 4.5 thinking 32k 63.7% $0.52
MiniMax-M2.5 63.7% $0.07
GPT-5.4 Mini xhigh 63.7% $0.51
Grok 4.7 low 63.3% $0.10
GPT-6 Luna medium 61% $0.006
o3 high 60.8% $0.50
GPT-5.6 Terra low 60.2% $0.09
o3-pro high 59.3% $4.16
o4-mini high 58.7% $0.41
Claude Opus 4.5 thinking, 8k 58.7% $0.38
Muse Spark Open low 58.5% $0.06
GPT-5.4 Mini high 58% $0.27
GPT-5.1 thinking, medium 57.7% $0.31
Gemini 3 Flash medium 57.7% $0.06
Claude Haiku 5.5 medium 57.3% $0.008
o3-pro medium 57% $3.18
DeepSeek-V3.2 57% $0.08
GPT-5.6 Luna medium 56.5% $0.06
GPT-5 medium 56.2% $0.33
GPT-5.2 low 55.7% $0.17
Beam 1.0 medium 55.7% $0.12
GPT-5 mini high 54.3% $0.12
GPT-5.6 Luna medium 54% $0.01
o3 medium 53.8% $0.29
Gemini 3.5 Flash-Lite high 53.5% $0.09
Inkling-Small low 52.5% $0.03
GPT-5.4 Nano xhigh 51.5% $0.13
Claude Haiku 5.5 low 50.3% $0.005
Gemini 3.5 Flash minimal 48.8% $0.06
Grok 4 Fast 48.5% $0.03
Inkling-Small minimal 48.5% $0.02
Claude Sonnet 4.5 thinking 16k 48.3% $0.27
Claude Haiku 4.5 thinking 32k 47.7% $0.26
GLM-5.3-Flash low 47% $0.02
Claude Sonnet 4.5 thinking 8k 46.5% $0.18
Beam 1.0 low 45% $0.08
GLM-5 44.7% $0.17
o3-pro low 44.3% $1.64
GPT-5 low 44% $0.15
o4-mini medium 41.8% $0.15
o3 low 41.5% $0.18
Gemini 2.5 Pro Jun 2025 thinking 16k 41% $0.48
GPT-5.4 Mini medium 40.8% $0.16
Claude Sonnet 4 thinking 16k 40% $0.37
tiny-recursion-model 40% $1.76
Claude Opus 4.5 thinking, none 40% $0.13
GPT-5.6 Luna low 39.5% $0.008
GPT-5.4 Nano high 38.2% $0.07
GPT-6 Luna low 37.7% $0.003
GPT-5 mini medium 37.3% $0.04
Claude Haiku 4.5 thinking 16k 37.3% $0.10
Gemini 2.5 Pro Jun 2025 thinking 32k 37% $0.51
Claude Opus 4 thinking 16k 35.7% $1.25
o3-mini high 34.5% $0.40
Gemini 3.6 Flash minimal 34.5% $0.04
GPT-5.6 Luna low 34.2% $0.04
Qwen 3.8 27B none 34% $0.04
Gemini 2.5 Flash May 2025 33.3% $0.04
Gemini 2.5 Flash May 2025 thinking 16k 33.3% $0.21
GPT-5.1 thinking, low 33.2% $0.10
GPT-5.4 Nano medium 33% $0.04
Gemini 2.5 Flash May 2025 thinking 24k 32.3% $0.20
Gemini 3.5 Flash-Lite medium 32.3% $0.03
Claude Sonnet 4.5 thinking 1k 31% $0.09
Claude Opus 4 thinking 8k 30.7% $0.74
Gemini 2.5 Pro Jun 2025 thinking 8k 29.5% $0.29
GPT-6 Sol none 29.3% $0.04
Claude Sonnet 4 thinking 8k 29% $0.20
Gemini 3 Flash low 29% $0.02
Claude 3.7 Sonnet 16k 28.6% $0.33
Claude Sonnet 4 thinking 1k 28% $0.09
codex-mini-2025-05-16 27.3% $0.16
Claude Opus 4 thinking 1k 27% $0.50
GPT-5 mini low 26.3% $0.01
Gemini 2.5 Flash May 2025 thinking 8k 25.8% $0.13
Claude Sonnet 4.5 25.5% $0.08
Claude Haiku 4.5 thinking 8k 25.5% $0.07
Claude Sonnet 4 23.8% $0.08
Nemotron 3.5 Super VL high 22.7% $0.03
Claude Opus 4 22.5% $0.40
o3-mini medium 22.3% $0.19
Gemini 3 Flash minimal 21.5% $0.01
o4-mini low 21.3% $0.04
DeepSeek-R1 May 2025 21.2% $0.05
Claude 3.7 Sonnet 8k 21.2% $0.21
GPT-5 nano medium 20.7% $0.01
Nemotron 3.5 Super VL medium 20.2% $0.01
GPT-5.4 Nano low 18.3% $0.01
Gemini 3.5 Flash-Lite low 17% $0.01
Claude Haiku 4.5 thinking 1k 16.8% $0.03
GPT-5 nano high 16.7% $0.03
Grok-3 mini low 16.5% $0.01
Gemini 2.5 Flash May 2025 thinking 1k 16% $0.04
Gemini 2.5 Pro Jun 2025 thinking 1k 16% $0.06
DeepSeek-R1 15.8% $0.06
o3-mini low 14.5% $0.05
Claude Haiku 4.5 14.3% $0.03
o1-mini 14% $0.14
Claude 3.7 Sonnet 13.6% $0.06
GPT-5.4 Mini low 13% $0.04
DeepSeek V4 Pro 0813 none 13% $0.02
GPT-5.2 12.3% $0.05
DeepSeek V4 Flash 0731 none 11.8% $0.002
Claude 3.7 Sonnet 1k 11.6% $0.07
Qwen3-235B-A22B-Instruct Jul 2025 11% $0.003
GPT-4.5 10.3% $0.29
GPT-6 Luna none 8.83% $0.002
DeepSeek V4.1 Flash none 7.83% $0.006
Gemini 3.5 Flash-Lite minimal 7.50% $0.008
Magistral Medium 1.0 thinking 6.12% $0.10
GPT-5 minimal 6.00% $0.03
Magistral Medium 1.0 5.91% $0.10
GPT-5.1 thinking, none 5.83% $0.03
DeepSeek V4.1 Flash 1017 none 5.67% $0.006
Grok 3 5.50% $0.09
GPT-4.1 5.50% $0.04
Inkling-Small none 5.50% $0.006
GPT-5 mini minimal 5.33% $0.006
Magistral Small 1.0 5.00% $0.04
GPT-4o Nov 2024 4.50% $0.05
Llama 4 Maverick 4.38% $0.008
GPT-5 nano low 4.04% $0.003
GPT-4.1 mini 3.50% $0.008
GPT-5.6 Luna none 3.33% $0.005
Nemotron 3.5 Super VL none 2.00% $0.004
GPT-5 nano minimal 1.50% $0.002
Llama 4 Scout 0.50% $0.004
GPT-4.1 nano 0% $0.002

The best score, 98.5%, costs $0.06 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI DeepSeek Z.ai Anthropic Inception Labs Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale GPT-6.1 Sol, high GPT-6.1 Sol, low DeepSeek V4 Flash 0731, max Mercury 3, medium GPT-6 Luna, high GPT-6 Luna, medium Claude Haiku 5.5, low 0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale GPT-6.1 Sol, high GPT-6.1 Sol, medium

Source: Official ARC-AGI leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5 max 98.5% $2.11
Claude Fable 5 xhigh 98.5% $1.02
Gemini 3.8 Flash high 98.5% $0.21
GPT-6 Astra xhigh 98.5% $0.35
GPT-6 Astra high 98.5% $0.28
Claude Opus 5.5 high 98.5% $0.16
GPT-6.1 Sol xhigh 98.5% $0.07
GPT-6.1 Sol high 98.5% $0.06
Gemini 4 Argon high 98.5% $0.16
Gemini 4 Argon medium 98.5% $0.06
Gemini 4 Carbon medium 98.5% $0.16
Claude Fable 5.2 xhigh 98.5% $0.36
GPT-6.2 Astra max 98.5% $0.35
GPT-6.2 Astra xhigh 98.5% $0.22
GPT-6.2 Astra high 98.5% $0.20
GPT-6.2 Astra medium 98.5% $0.17
GPT-6.2 Astra low 98.5% $0.17
Gemini 3.1 Pro 98% $0.52
GPT-5.6 Sol xhigh 97.5% $0.40
Claude Opus 5 high 97.5% $0.47
Claude Opus 5 max 97.5% $0.70
Claude Fable 5.1 max 97.5% $1.40
Gemini 3.8 Flash medium 97.5% $0.11
GPT-6 Astra max 97.5% $0.43
GPT-6 Astra medium 97.5% $0.24
Claude Opus 5.5 max 97.5% $0.64
Claude Opus 5.5 xhigh 97.5% $0.22
Claude Opus 5.5 medium 97.5% $0.15
Gemini 4 Carbon high 97.5% $0.41
Claude Fable 5.2 max 97.5% $0.92
Claude Fable 5.2 high 97.5% $0.33
Claude Fable 5.2 medium 97.5% $0.26
GPT-6.2 Sol max 97.5% $0.11
GPT-6.2 Sol high 97.5% $0.06
GPT-5.6 Sol high 97% $0.30
GPT-6.2 Sol xhigh 97% $0.08
GPT-5.5 Pro high 96.5% $4.53
GPT-5.6 Sol max 96.5% $0.54
GPT-5.6 Terra max 96.5% $0.55
Claude Fable 5.1 xhigh 96.5% $0.96
GPT-6 Astra low 96.5% $0.23
Claude Sonnet 5.5 max 96.5% $0.50
Claude Sonnet 5.5 xhigh 96.5% $0.14
GPT-6.1 Sol max 96.5% $0.10
Inkling 2 Small xhigh 96.5% $0.08
Gemini 3 Deep Think 96% $7.17
Claude Fable 5.1 high 96% $0.52
Claude Fable 5 high 95.5% $0.77
Gemini 3.7 Flash high 95.5% $0.12
GPT-6 Sol max 95.5% $0.14
GPT-6.1 Sol medium 95.5% $0.05
GPT-5.5 xhigh 95% $0.73
GPT-5.5 Pro xhigh 95% $4.52
GPT-5.4 Pro xhigh 94.5% $7.21
GPT-5.5 high 94.5% $0.56
Kimi K3 max 94.5% $0.77
Claude Fable 5.1 medium 94.5% $0.44
DeepSeek V4.1 Flash max 94.5% $0.07
DeepSeek V4.1 Flash 1017 high 94.5% $0.06
Claude Fable 5.2 low 94.5% $0.23
GPT-6.2 Sol medium 94.5% $0.05
Claude Opus 4.6 120k, high 94% $1.40
GPT-5.6 Terra xhigh 94% $0.26
GPT-5.4 xhigh 93.7% $0.62
Claude Opus 4.7 high 93.5% $1.41
GPT-6.1 Sol low 93.5% $0.05
DeepSeek V4.1 Flash 1017 max 93.5% $0.06
Claude Opus 4.6 120k, max 93% $1.88
GPT-5.4 high 92.7% $0.37
GPT-6 Sol xhigh 92.7% $0.10
Gemini 3.5 Flash high 92.5% $0.43
Claude Opus 4.8 max 92.5% $2.33
Claude Fable 5 medium 92.5% $0.66
GPT-5.6 Sol medium 92.5% $0.22
GPT-5.5 medium 92.2% $0.39
Claude Opus 4.6 120k, medium 92% $0.96
Claude Opus 4.7 max 92% $2.58
Claude Opus 4.8 high 92% $1.04
GPT-5.6 Terra high 92% $0.19
Claude Sonnet 5.5 high 92% $0.11
Claude Opus 4.8 medium 91.5% $0.91
DeepSeek V4.1 Flash 1017 low 91.5% $0.05
Gemini 3.6 Flash high 91.2% $0.34
Gemini 3.7 Flash medium 91.2% $0.06
Claude Opus 4.7 low 91% $0.76
Claude Opus 4.7 medium 91% $1.04
GLM-5.3-Flash max 91% $0.04
GPT-6 Sol high 91% $0.08
GPT-5.6 Luna max 90.7% $0.07
GPT-5.2 Pro xhigh 90.5% $12
Claude Fable 5 low 90.5% $0.51
DeepSeek V4 Pro 0813 low 90.5% $0.18
Gemini 3.8 Flash low 90.5% $0.07
DeepSeek V4.1 Flash low 90.5% $0.05
Grok 4.7 high 90.2% $0.64
DeepSeek V4 Pro 0813 max 90% $0.30
Claude Fable 5.1 low 90% $0.37
Grok 4.7 medium 90% $0.52
Grok 4.20 reasoning 89.5% $0.35
Grok 4.7 xhigh 89.5% $0.78
DeepSeek V4 Flash 0731 max 89% $0.02
Claude Haiku 5.5 max 89% $0.05
Muse Spark Open xhigh 89% $0.29
DeepSeek V4.1 Flash high 88.5% $0.06
Claude Opus 5.5 low 88.5% $0.12
Muse Spark Open high 88.5% $0.23
Mercury 3 high 88.5% $0.03
Claude Opus 4.8 low 88% $0.67
GPT-5.6 Luna max 88% $0.32
GPT-5.6 Luna xhigh 87.7% $0.21
Gemini 3 Deep Think 87.5% $44
Grok 4.6 medium 87.5% $0.30
Qwen 3.8 27B xhigh 87.5% $0.22
Claude Haiku 5.5 xhigh 87.5% $0.03
Grok 4.5 medium 87.2% $0.31
DeepSeek V4 Pro 0813 high 87.2% $0.31
DeepSeek V4 Flash 0731 high 87% $0.02
Grok 4.6 xhigh 87% $0.35
Grok 4.6 high 87% $0.33
Kimi K3 high 86.7% $0.48
GPT-6 Luna max 86.7% $0.02
Claude Sonnet 4.6 high 86.5% $1.19
GPT-5.2 xhigh 86.2% $0.96
GPT-5.4 medium 86.2% $0.25
Claude Opus 4.6 120k, low 86% $0.65
Claude Sonnet 4.6 max 86% $1.45
GPT-6 Astra none 86% $0.21
GPT-5.2 Pro high 85.7% $5.87
Grok 4.5 high 85.7% $0.33
Gemini 4 Carbon low 85.5% $0.06
Gemini 3.7 Flash low 85.2% $0.04
GPT-5.6 Luna xhigh 85% $0.05
Gemini 3 Flash high 84.7% $0.17
Mercury 3 medium 84.7% $0.02
Inkling-Small xhigh 84% $0.12
DeepSeek V4 Flash 0731 low 84% $0.01
GPT-6.2 Sol low 84% $0.04
GPT-6 Sol medium 83.7% $0.07
Gemini 3.6 Flash medium 83.2% $0.25
Claude Haiku 5.5 high 81.5% $0.02
GPT-5.2 Pro medium 81.2% $3.98
Mercury 3 low 80.7% $0.02
Claude Opus 4.5 thinking, 64k 80% $1.47
Inkling 79.5% $0.30
Grok 4.5 low 79.2% $0.18
GPT-5.2 high 78.7% $0.52
GPT-5.6 Luna high 78% $0.03
Inkling-Small high 78% $0.07
Muse Spark Open medium 77.7% $0.12
GLM-5.2 77% $0.19
GPT-5.6 Terra medium 77% $0.13
GPT-5.6 Luna high 76.5% $0.13
Gemini 3.6 Flash low 76.5% $0.12
Gemini 4 Argon low 76.5% $0.04
GPT-5.5 low 76.2% $0.20
Claude Opus 4.5 thinking, 32k 75.8% $0.95
Gemini 3 Pro 75% $0.49
Grok 4.6 low 74.8% $0.10
GPT-5.6 Sol low 74.5% $0.17
Beam 1.0 xhigh 74% $0.31
GPT-6 Luna xhigh 73% $0.01
GPT-5.1 thinking, high 72.8% $0.67
GPT-5.2 medium 72.7% $0.34
GPT-6 Sol low 72.2% $0.05
Claude Opus 4.5 thinking, 16k 72% $0.62
GLM-5.3-Flash high 71.8% $0.03
GPT-6 Luna high 70.3% $0.009
GPT-5 Pro 70.2% $4.78
Claude Sonnet 5.5 medium 70% $0.07
Qwen 3.8 27B low 69.2% $0.12
Qwen 3.8 27B medium 68.7% $0.13
GPT-5.4 low 68.2% $0.15
Beam 1.0 high 67.7% $0.20
Inkling-Small medium 67% $0.03
Claude Sonnet 5.5 low 67% $0.06
Grok 4 thinking 66.7% $1.01
GPT-5 high 65.7% $0.51
Kimi K3 low 65.7% $0.18
Kimi K2.5 65.3% $0.15
Claude Sonnet 4.5 thinking 32k 63.7% $0.52
MiniMax-M2.5 63.7% $0.07
GPT-5.4 Mini xhigh 63.7% $0.51
Grok 4.7 low 63.3% $0.10
GPT-6 Luna medium 61% $0.006
o3 high 60.8% $0.50
GPT-5.6 Terra low 60.2% $0.09
o3-pro high 59.3% $4.16
o4-mini high 58.7% $0.41
Claude Opus 4.5 thinking, 8k 58.7% $0.38
Muse Spark Open low 58.5% $0.06
GPT-5.4 Mini high 58% $0.27
GPT-5.1 thinking, medium 57.7% $0.31
Gemini 3 Flash medium 57.7% $0.06
Claude Haiku 5.5 medium 57.3% $0.008
o3-pro medium 57% $3.18
DeepSeek-V3.2 57% $0.08
GPT-5.6 Luna medium 56.5% $0.06
GPT-5 medium 56.2% $0.33
GPT-5.2 low 55.7% $0.17
Beam 1.0 medium 55.7% $0.12
GPT-5 mini high 54.3% $0.12
GPT-5.6 Luna medium 54% $0.01
o3 medium 53.8% $0.29
Gemini 3.5 Flash-Lite high 53.5% $0.09
Inkling-Small low 52.5% $0.03
GPT-5.4 Nano xhigh 51.5% $0.13
Claude Haiku 5.5 low 50.3% $0.005
Gemini 3.5 Flash minimal 48.8% $0.06
Grok 4 Fast 48.5% $0.03
Inkling-Small minimal 48.5% $0.02
Claude Sonnet 4.5 thinking 16k 48.3% $0.27
Claude Haiku 4.5 thinking 32k 47.7% $0.26
GLM-5.3-Flash low 47% $0.02
Claude Sonnet 4.5 thinking 8k 46.5% $0.18
Beam 1.0 low 45% $0.08
GLM-5 44.7% $0.17
o3-pro low 44.3% $1.64
GPT-5 low 44% $0.15
o4-mini medium 41.8% $0.15
o3 low 41.5% $0.18
Gemini 2.5 Pro Jun 2025 thinking 16k 41% $0.48
GPT-5.4 Mini medium 40.8% $0.16
Claude Sonnet 4 thinking 16k 40% $0.37
tiny-recursion-model 40% $1.76
Claude Opus 4.5 thinking, none 40% $0.13
GPT-5.6 Luna low 39.5% $0.008
GPT-5.4 Nano high 38.2% $0.07
GPT-6 Luna low 37.7% $0.003
GPT-5 mini medium 37.3% $0.04
Claude Haiku 4.5 thinking 16k 37.3% $0.10
Gemini 2.5 Pro Jun 2025 thinking 32k 37% $0.51
Claude Opus 4 thinking 16k 35.7% $1.25
o3-mini high 34.5% $0.40
Gemini 3.6 Flash minimal 34.5% $0.04
GPT-5.6 Luna low 34.2% $0.04
Qwen 3.8 27B none 34% $0.04
Gemini 2.5 Flash May 2025 33.3% $0.04
Gemini 2.5 Flash May 2025 thinking 16k 33.3% $0.21
GPT-5.1 thinking, low 33.2% $0.10
GPT-5.4 Nano medium 33% $0.04
Gemini 2.5 Flash May 2025 thinking 24k 32.3% $0.20
Gemini 3.5 Flash-Lite medium 32.3% $0.03
Claude Sonnet 4.5 thinking 1k 31% $0.09
Claude Opus 4 thinking 8k 30.7% $0.74
Gemini 2.5 Pro Jun 2025 thinking 8k 29.5% $0.29
GPT-6 Sol none 29.3% $0.04
Claude Sonnet 4 thinking 8k 29% $0.20
Gemini 3 Flash low 29% $0.02
Claude 3.7 Sonnet 16k 28.6% $0.33
Claude Sonnet 4 thinking 1k 28% $0.09
codex-mini-2025-05-16 27.3% $0.16
Claude Opus 4 thinking 1k 27% $0.50
GPT-5 mini low 26.3% $0.01
Gemini 2.5 Flash May 2025 thinking 8k 25.8% $0.13
Claude Sonnet 4.5 25.5% $0.08
Claude Haiku 4.5 thinking 8k 25.5% $0.07
Claude Sonnet 4 23.8% $0.08
Nemotron 3.5 Super VL high 22.7% $0.03
Claude Opus 4 22.5% $0.40
o3-mini medium 22.3% $0.19
Gemini 3 Flash minimal 21.5% $0.01
o4-mini low 21.3% $0.04
DeepSeek-R1 May 2025 21.2% $0.05
Claude 3.7 Sonnet 8k 21.2% $0.21
GPT-5 nano medium 20.7% $0.01
Nemotron 3.5 Super VL medium 20.2% $0.01
GPT-5.4 Nano low 18.3% $0.01
Gemini 3.5 Flash-Lite low 17% $0.01
Claude Haiku 4.5 thinking 1k 16.8% $0.03
GPT-5 nano high 16.7% $0.03
Grok-3 mini low 16.5% $0.01
Gemini 2.5 Flash May 2025 thinking 1k 16% $0.04
Gemini 2.5 Pro Jun 2025 thinking 1k 16% $0.06
DeepSeek-R1 15.8% $0.06
o3-mini low 14.5% $0.05
Claude Haiku 4.5 14.3% $0.03
o1-mini 14% $0.14
Claude 3.7 Sonnet 13.6% $0.06
GPT-5.4 Mini low 13% $0.04
DeepSeek V4 Pro 0813 none 13% $0.02
GPT-5.2 12.3% $0.05
DeepSeek V4 Flash 0731 none 11.8% $0.002
Claude 3.7 Sonnet 1k 11.6% $0.07
Qwen3-235B-A22B-Instruct Jul 2025 11% $0.003
GPT-4.5 10.3% $0.29
GPT-6 Luna none 8.83% $0.002
DeepSeek V4.1 Flash none 7.83% $0.006
Gemini 3.5 Flash-Lite minimal 7.50% $0.008
Magistral Medium 1.0 thinking 6.12% $0.10
GPT-5 minimal 6.00% $0.03
Magistral Medium 1.0 5.91% $0.10
GPT-5.1 thinking, none 5.83% $0.03
DeepSeek V4.1 Flash 1017 none 5.67% $0.006
Grok 3 5.50% $0.09
GPT-4.1 5.50% $0.04
Inkling-Small none 5.50% $0.006
GPT-5 mini minimal 5.33% $0.006
Magistral Small 1.0 5.00% $0.04
GPT-4o Nov 2024 4.50% $0.05
Llama 4 Maverick 4.38% $0.008
GPT-5 nano low 4.04% $0.003
GPT-4.1 mini 3.50% $0.008
GPT-5.6 Luna none 3.33% $0.005
Nemotron 3.5 Super VL none 2.00% $0.004
GPT-5 nano minimal 1.50% $0.002
Llama 4 Scout 0.50% $0.004
GPT-4.1 nano 0% $0.002

14 models have held number one on the official leaderboard since 2023 15 models have held number one with claims included since 2023

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 DeepSeek xAI Google DeepMind Anthropic ByteDance Seed Meituan Ant Group
Own leaderboard Passed 85% in Dec 2025, by GPT-5.2 Pro
Lab claims Passed 85% in Dec 2025, by GPT-5.2 Pro

Sources: Official ARC-AGI leaderboard and labs’ model cards

GPT-5.2 Pro on the official leaderboard passed 85% in December 2025

That was 73 months after its release. 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 A claim Epoch AI collected from a report or paper Best so far
0% 20% 40% 60% 80% 100% 2025 2026 2027 Near the ceiling (85%) Claude Fable 5 Claude Opus 4.6 Claude Opus 4.5 Gemini 3 Pro GPT-5 Pro o3 Seed-Thinking-v1.5 0% 20% 40% 60% 80% 100% 2025 2026 2027 Near the ceiling (85%) Claude Fable 5 GPT-5.2 Pro

Sources: Official ARC-AGI leaderboard, labs’ model cards and claims collected by Epoch AI (CC BY 4.0)

2 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.

ByteDance Seed
Meituan
Ant Group
Tencent
OpenAI
Anthropic
Thinking Machines
DateLaunch postScoreAs the post writes it
Nov 20, 2026 Inkling 2 Mini 84.5 ARC-AGI v1
Sep 1, 2026 Claude Fable 5.1 97.5 ARC-AGI-1
Jul 30, 2026 Inkling-Small 84 ARC-AGI-1
Jul 24, 2026 Claude Opus 5 97.5 ARC-AGI-1
Apr 23, 2026 GPT-5.5 95 ARC-AGI-1 (Verified)
Apr 16, 2026 Claude Opus 4.7 92 ARC-AGI-1
Mar 5, 2026 GPT-5.4 Pro 94.5 ARC-AGI-1 (Verified)
Mar 5, 2026 GPT-5.4 93.7 ARC-AGI-1 (Verified)
Feb 14, 2026 Seed2.0 Pro 85.4 ARC-AGI-1 (text) 85.4 / ArcAGI1-Image 88.8
Dec 18, 2025 Seed1.8 67.9 ARC-AGI-1
Dec 11, 2025 GPT-5.2 Pro 90.5 ARC-AGI-1 (Verified)
Dec 11, 2025 GPT-5.2 86.2 ARC-AGI-1 (Verified) Abstract reasoning
Dec 5, 2025 Tencent HY 2.0 Think 52.9 ARC AGI_1
Oct 10, 2025 Ring-1T 55.9 ARC-AGI-1 (pass@1)
Sep 21, 2025 LongCat-Flash-Thinking 50.3 ARC-AGI(Mean@1)
Apr 10, 2025 Seed-Thinking-v1.5 39.9 ARC-AGI

16 system cards report it, from 7 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 84.5 ARC-AGI v1 Capabilities
Oct 19, 2026 Muse Glimmer technical report 66.9 ARC-AGI, 400 questions, pass@1 over 4 attempts, high, 67.9 xhigh (Table 5) Capabilities
Sep 1, 2026 Claude Fable 5.1 system card 97.5 ARC-AGI-1 semi-private set, max effort, verified by ARC Prize Foundation (Table 8.1.A) Capabilities
Jul 30, 2026 Inkling-Small model card 84 ARC-AGI-1 (5. Evaluations table) Capabilities
Jul 24, 2026 Claude Opus 5 system card 97.5 ARC-AGI-1 semi-private set, max effort, verified by ARC Prize Foundation (97.50%) (Table 8.1.A, 8.14.1) Capabilities
Jun 13, 2026 Ling-2.6-1T technical report 50.9 ARCPrize, Mean@4, cited as the ARC Prize 2024 technical report (Table 5) Capabilities
Apr 16, 2026 Claude Opus 4.7 system card 92 ARC-AGI-1 (Table 8.1.A). Section 8.11 gives 93.5 at high effort Capabilities
Mar 21, 2026 Seed1.8 model card 67.9 ARC-AGI-1 (Table 1) Capabilities
Feb 17, 2026 Claude Sonnet 4.6 system card 86.5 ARC-AGI-1, 120k thinking tokens, high effort, ARC Prize private set (2.7, Figure 2.7.A) Capabilities
Feb 14, 2026 Seed2.0 Pro model card 85.4 ARC-AGI-1, text input (Table 3). ArcAGI1-Image with text and image input 88.8 (Table 8) Capabilities
Feb 11, 2026 Step 3.5 Flash technical report 54.8 ARC-AGI-1 (Table 12). With Python 56.5 Capabilities
Feb 5, 2026 Claude Opus 4.6 system card 94 ARC-AGI-1, 120k thinking tokens, high effort, ARC Prize private set, reported by ARC Prize Foundation (2.9, Figure 2.9) Capabilities
Nov 24, 2025 Claude Opus 4.5 system card 80 ARC-AGI-1, semi-private set, 64k thinking, reported by ARC Prize Foundation (2.10, Figure 2.10.A) Capabilities
Oct 21, 2025 Ring-1T technical report 55.9 ARC-AGI-1 (Table 1) Capabilities
Sep 23, 2025 LongCat-Flash-Thinking technical report 50.3 ARC-AGI(Mean@1) (Table 2) Capabilities
Apr 10, 2025 Seed-Thinking-v1.5 technical report 39.9 ARC-AGI (Table 2) Capabilities

12 lab claims next to the leaderboard

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. 11 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
Claude Opus 4.7 92% ARC-AGI-1 93.5% Leaderboard, high −1.5 pts
GPT-5.2 86.2% ARC-AGI-1 (Verified) Abstract reasoning 86.2% Leaderboard, xhigh Same
GPT-5.4 93.7% ARC-AGI-1 (Verified) 93.7% Leaderboard, xhigh Same
GPT-5.2 Pro 90.5% ARC-AGI-1 (Verified) 90.5% Leaderboard, xhigh Same
GPT-5.4 Pro 94.5% ARC-AGI-1 (Verified) 94.5% Leaderboard, xhigh Same
GPT-5.5 95% ARC-AGI-1 (Verified) 95% Leaderboard, xhigh Same
Claude Opus 5 97.5% ARC-AGI-1 97.5% Leaderboard, high Same
Inkling-Small 84% ARC-AGI-1 84% Leaderboard, xhigh Same
Claude Fable 5.1 97.5% ARC-AGI-1 97.5% Leaderboard, max Same
Claude Opus 4.5 80% ARC-AGI-1, semi-private set, 64k thinking, reported by ARC Prize Foundation (2.10, Figure 2.10.A) 80% Leaderboard, thinking, 64k Same
Claude Opus 4.6 94% ARC-AGI-1, 120k thinking tokens, high effort, ARC Prize private set, reported by ARC Prize Foundation (2.9, Figure 2.9) 94% Leaderboard, 120k, high Same
Claude Sonnet 4.6 86.5% ARC-AGI-1, 120k thinking tokens, high effort, ARC Prize private set (2.7, Figure 2.7.A) 86.5% Leaderboard, high Same

One logged change that moves scores

Changes by the maintainers, so a score from before a change may not compare with one from after it. Newest first, each with the maintainers' own source. The list may be incomplete. What we log.

  1. Questions corrected June 21st, 2024

    François Chollet released corrections to two training tasks and nine evaluation tasks in ARC-AGI 1.0.2. Earlier copies contain different examples or expected outputs, so scores should identify the dataset revision. Source

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.

OpenAI DeepSeek xAI Google DeepMind Anthropic
ARC-AGI-3 Passed 85% in Sep 2026, by GPT-6 Astra
ARC-AGI Passed 85% in Dec 2025, by GPT-5.2 Pro
ARC-AGI-2 Passed 85% in Apr 2026, by GPT-5.5
ARC-AGI-3 Passed 85% in Sep 2026, by GPT-6 Astra

Source: Official leaderboards

More in Science and reasoning

All 204 in Science and reasoning
  • GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics. Gemini 4 Carbon 97%
  • Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
  • BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Fable 5.2 83.7%
  • FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics. Intern-S2-397B 87%
  • HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels. Gemini 3 Pro 78.3%
  • LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%

Data audit

All 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (101 entries), launch posts (16 scores), system cards (16 scores) and claims Epoch AI collected (30 models).
  • Pass Models scored 109 models we track, plus 3 other entries on its leaderboard.
  • Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
  • Pass Frontier models 9 of the top 10 have a score, led by GPT-6.2 Astra (#1).
  • Note Headroom Saturated. The best is 98.5%, past 85% since December 2025.
Leaderboard import 3 pass
  • Pass Importer We read the board on arcprize.org: 307 rows for models we track and 3 other entries at the last read.
  • Pass Last read November 18th, 2026.
  • Pass Read worked The last read worked.
Consistency 2 pass
  • Pass Lab claims All 12 pairs agree within 10 points.
  • Pass Known issues No published audit found problems.

How we check