Benchmark Almanac

Mystery Game Puzzles

Claude Fable 5.2 90% leads since November 2026
Not in model cards

100 mid-game positions from an undisclosed puzzle variant of a well-known game. Model picks the single best move, checked against an answer key.

Game identity, prompts and positions are kept secret to prevent targeted preparation. Likely the 'game-puzzles' benchmark in the Epoch post.

Made by
Epoch AI
Released
July 2026
Kind
Puzzles and chess
Level
Competition
Scored
Checked automatically
Answers
Short answer
Tasks
Kept private
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last model added
November 2026
Scores refreshed
November 12th, 2026
84
Models scored in Epoch AI’s runs

Claude Fable 5.2 leads with 90%

Best score per model from Epoch AI’s own runs. Where a model ran with several settings or harnesses, we keep its best.

84 models. Scroll sideways for the rest.

  1. 90.0 Claude Fable 5.2
  2. 84.0 GPT-6 Astra
  3. 80.0 GPT-6.1 Sol
  4. 74.0 GPT-6.1 Terra
  5. 73.0 Gemini 4 Carbon
  6. 71.0 Claude Opus 5.5
  7. 65.0 Claude Sonnet 5.5
  8. 64.0 Gemini 4 Argon
  9. 59.0 Claude Opus 5
  10. 58.0 GPT-5.6 Sol
  11. 58.0 Claude Fable 5.1
  12. 56.0 GPT-5.5
  13. 56.0 GPT-6 Sol
  14. 52.0 Claude Fable 5
  15. 51.0 Muse Wonder
  16. 49.0 DeepSeek V4.1 Flash 1017
  17. 47.0 Gemini 3.8 Flash
  18. 43.0 DeepSeek V4 Pro 0813
  19. 43.0 DeepSeek V4.1 Flash
  20. 42.0 Kimi K3.1 Pro
  21. 38.0 Qwen 3.8 Max
  22. 37.0 GPT-5.4
  23. 37.0 Gemini 3.7 Flash
  24. 36.0 Claude Opus 4.8
  25. 35.0 Claude Sonnet 5
  26. 35.0 GPT-5.6 Terra
  27. 34.0 Gemini 3.1 Pro
  28. 34.0 DeepSeek V4 Flash 0731
  29. 34.0 Grok 4.6
  30. 33.0 GLM-5.3
  31. 32.0 Gemini 3.5 Flash
  32. 32.0 Qwen3.7-Max
  33. 30.0 Gemini 3.6 Flash
  34. 30.0 Claude Haiku 5.5
  35. 29.0 o3
  36. 29.0 Grok 4.7
  37. 28.0 Claude Opus 4.7
  38. 26.0 Gemini 3 Flash
  39. 26.0 Kimi K3
  40. 25.0 Claude Opus 4.6
  41. 25.0 Muse Spark 1.3
  42. 23.0 GPT-5
  43. 23.0 GPT-5.2
  44. 22.0 Claude Opus 4.5
  45. 22.0 Qwen 3.6 35B-A3B
  46. 21.0 Claude Opus 4.1
  47. 21.0 GPT-5.6 Luna
  48. 20.0 Qwen 3.5 Flash (hosted 35B-A3B)
  49. 20.0 Nemotron 3 Ultra
  50. 20.0 Mistral Large 4
  51. 19.0 GPT-5.1
  52. 19.0 Qwen 3.6 Max (Preview)
  53. 19.0 GLM-5.2
  54. 19.0 Gemini 3.5 Flash-Lite
  55. 18.0 Qwen3.5 397B-A17B
  56. 18.0 Kimi K2.6
  57. 18.0 Qwen 3.6 Flash
  58. 17.0 Claude Sonnet 4.5
  59. 17.0 Qwen 3.5 Plus (hosted 397B-A17B)
  60. 17.0 Qwen3.5-122B-A10B
  61. 17.0 DeepSeek-V4-Pro
  62. 17.0 Qwen3.7-Plus
  63. 17.0 Mistral Large 4 Preview
  64. 16.0 Claude Sonnet 4.6
  65. 15.0 Qwen3.7 Flash
  66. 12.0 GPT-4
  67. 12.0 GPT-4o mini
  68. 12.0 Qwen 3.6 Plus
  69. 11.0 GPT-5.4 Mini
  70. 10.0 GPT-5 mini
  71. 9.00 Qwen3-235B-A22B-Thinking
  72. 9.00 GPT-5 nano
  73. 9.00 GPT-5.4 Nano
  74. 8.00 MiniMax-M3
  75. 8.00 GLM-5.3-Flash
  76. 7.00 o3-mini
  77. 7.00 GPT-4.1 mini
  78. 7.00 Qwen3.6 27B
  79. 7.00 GPT-6 Luna
  80. 6.00 Inkling-Small
  81. 5.00 o4-mini
  82. 5.00 Qwen3-Max
  83. 3.00 GPT-3.5 Turbo
  84. 2.00 gpt-oss-120b
  1. 1 Claude Fable 5.2 max 90%
  2. 2 GPT-6 Astra max 84%
  3. 3 GPT-6.1 Sol max 80%
  4. 4 GPT-6.1 Terra max 74%
  5. 5 Gemini 4 Carbon high 73%
  6. 6 Claude Opus 5.5 max 71%
  7. 7 Claude Sonnet 5.5 max 65%
  8. 8 Gemini 4 Argon high 64%
  9. 9 Claude Opus 5 max 59%
  10. 10 GPT-5.6 Sol max 58%
  11. 11 Claude Fable 5.1 max 58%
  12. 12 GPT-5.5 xhigh 56%
  13. 13 GPT-6 Sol max 56%
  14. 14 Claude Fable 5 max 52%
  15. 15 Muse Wonder max 51%
  16. 16 DeepSeek V4.1 Flash 1017 max 49%
  17. 17 Gemini 3.8 Flash high 47%
  18. 18 DeepSeek V4 Pro 0813 max 43%
  19. 19 DeepSeek V4.1 Flash max 43%
  20. 20 Kimi K3.1 Pro Together, max 42%
  21. 21 Qwen 3.8 Max xhigh 38%
  22. 22 GPT-5.4 xhigh 37%
  23. 23 Gemini 3.7 Flash high 37%
  24. 24 Claude Opus 4.8 max 36%
  25. 25 Claude Sonnet 5 max 35%
  26. 26 GPT-5.6 Terra max 35%
  27. 27 Gemini 3.1 Pro high 34%
  28. 28 DeepSeek V4 Flash 0731 max 34%
  29. 29 Grok 4.6 xhigh 34%
  30. 30 GLM-5.3 max 33%
  31. 31 Gemini 3.5 Flash high 32%
  32. 32 Qwen3.7-Max 32%
  33. 33 Gemini 3.6 Flash high 30%
  34. 34 Claude Haiku 5.5 max 30%
  35. 35 o3 high 29%
  36. 36 Grok 4.7 xhigh 29%
  37. 37 Claude Opus 4.7 max 28%
  38. 38 Gemini 3 Flash low 26%
  39. 39 Kimi K3 max 26%
  40. 40 Claude Opus 4.6 max 25%
  41. 41 Muse Spark 1.3 max 25%
  42. 42 GPT-5 high 23%
  43. 43 GPT-5.2 high 23%
  44. 44 Claude Opus 4.5 48k thinking 22%
  45. 45 Qwen 3.6 35B-A3B none 22%
  46. 46 Claude Opus 4.1 24K 21%
  47. 47 GPT-5.6 Luna max 21%
  48. 48 Qwen 3.5 Flash (hosted 35B-A3B) 20%
  49. 49 Nemotron 3 Ultra 20%
  50. 50 Mistral Large 4 high 20%
  51. 51 GPT-5.1 low 19%
  52. 52 Qwen 3.6 Max (Preview) none 19%
  53. 53 GLM-5.2 low 19%
  54. 54 Gemini 3.5 Flash-Lite low 19%
  55. 55 Qwen3.5 397B-A17B none 18%
  56. 56 Kimi K2.6 18%
  57. 57 Qwen 3.6 Flash 2026-04-16 18%
  58. 58 Claude Sonnet 4.5 48k thinking 17%
  59. 59 Qwen 3.5 Plus (hosted 397B-A17B) none 17%
  60. 60 Qwen3.5-122B-A10B none 17%
  61. 61 DeepSeek-V4-Pro none 17%
  62. 62 Qwen3.7-Plus 17%
  63. 63 Mistral Large 4 Preview high 17%
  64. 64 Claude Sonnet 4.6 low 16%
  65. 65 Qwen3.7 Flash 15%
  66. 66 GPT-4 Jun 2023 12%
  67. 67 GPT-4o mini 12%
  68. 68 Qwen 3.6 Plus 2026-04-02 12%
  69. 69 GPT-5.4 Mini none 11%
  70. 70 GPT-5 mini minimal 10%
  71. 71 Qwen3-235B-A22B-Thinking Jul 2025 9.00%
  72. 72 GPT-5 nano medium 9.00%
  73. 73 GPT-5.4 Nano none 9.00%
  74. 74 MiniMax-M3 none 8.00%
  75. 75 GLM-5.3-Flash max 8.00%
  76. 76 o3-mini high 7.00%
  77. 77 GPT-4.1 mini 7.00%
  78. 78 Qwen3.6 27B none 7.00%
  79. 79 GPT-6 Luna max 7.00%
  80. 80 Inkling-Small xhigh 6.00%
  81. 81 o4-mini high 5.00%
  82. 82 Qwen3-Max 5.00%
  83. 83 GPT-3.5 Turbo Jan 2024 3.00%
  84. 84 gpt-oss-120b medium 2.00%

Source: Epoch AI runs (CC BY 4.0)

#ModelScoreSettingDateSource
1 Claude Fable 5.2 90% max Nov 2026 Epoch AI run
2 GPT-6 Astra 84% max Sep 2026 Epoch AI run
3 GPT-6.1 Sol 80% max Sep 2026 Epoch AI run
4 GPT-6.1 Terra 74% max Oct 2026 Epoch AI run
5 Gemini 4 Carbon 73% high Oct 2026 Epoch AI run
6 Claude Opus 5.5 71% max Sep 2026 Epoch AI run
7 Claude Sonnet 5.5 65% max Sep 2026 Epoch AI run
8 Gemini 4 Argon 64% high Sep 2026 Epoch AI run
9 Claude Opus 5 59% max Jul 2026 Epoch AI run
10 GPT-5.6 Sol 58% max Jul 2026 Epoch AI run
11 Claude Fable 5.1 58% max Sep 2026 Epoch AI run
12 GPT-5.5 56% xhigh Apr 2026 Epoch AI run
13 GPT-6 Sol 56% max Sep 2026 Epoch AI run
14 Claude Fable 5 52% max Jun 2026 Epoch AI run
15 Muse Wonder 51% max Oct 2026 Epoch AI run
16 DeepSeek V4.1 Flash 1017 49% max Oct 2026 Epoch AI run
17 Gemini 3.8 Flash 47% high Sep 2026 Epoch AI run
18 DeepSeek V4 Pro 0813 43% max Aug 2026 Epoch AI run
19 DeepSeek V4.1 Flash 43% max Sep 2026 Epoch AI run
20 Kimi K3.1 Pro 42% Together, max Oct 2026 Epoch AI run
21 Qwen 3.8 Max 38% xhigh Aug 2026 Epoch AI run
22 GPT-5.4 37% xhigh Mar 2026 Epoch AI run
23 Gemini 3.7 Flash 37% high Aug 2026 Epoch AI run
24 Claude Opus 4.8 36% max May 2026 Epoch AI run
25 Claude Sonnet 5 35% max Jun 2026 Epoch AI run
26 GPT-5.6 Terra 35% max Jul 2026 Epoch AI run
27 Gemini 3.1 Pro 34% high Feb 2026 Epoch AI run
28 DeepSeek V4 Flash 0731 34% max Jul 2026 Epoch AI run
29 Grok 4.6 34% xhigh Aug 2026 Epoch AI run
30 GLM-5.3 33% max Aug 2026 Epoch AI run
31 Gemini 3.5 Flash 32% high May 2026 Epoch AI run
32 Qwen3.7-Max 32% May 2026 Epoch AI run
33 Gemini 3.6 Flash 30% high Jul 2026 Epoch AI run
34 Claude Haiku 5.5 30% max Oct 2026 Epoch AI run
35 o3 29% high Apr 2025 Epoch AI run
36 Grok 4.7 29% xhigh Sep 2026 Epoch AI run
37 Claude Opus 4.7 28% max Apr 2026 Epoch AI run
38 Gemini 3 Flash 26% low Dec 2025 Epoch AI run
39 Kimi K3 26% max Jul 2026 Epoch AI run
40 Claude Opus 4.6 25% max Feb 2026 Epoch AI run
41 Muse Spark 1.3 25% max Sep 2026 Epoch AI run
42 GPT-5 23% high Aug 2025 Epoch AI run
43 GPT-5.2 23% high Dec 2025 Epoch AI run
44 Claude Opus 4.5 22% 48k thinking Nov 2025 Epoch AI run
45 Qwen 3.6 35B-A3B 22% none Apr 2026 Epoch AI run
46 Claude Opus 4.1 21% 24K Aug 2025 Epoch AI run
47 GPT-5.6 Luna 21% max Jul 2026 Epoch AI run
48 Qwen 3.5 Flash (hosted 35B-A3B) 20% Feb 2026 Epoch AI run
49 Nemotron 3 Ultra 20% Jun 2026 Epoch AI run
50 Mistral Large 4 20% high Oct 2026 Epoch AI run
51 GPT-5.1 19% low Nov 2025 Epoch AI run
52 Qwen 3.6 Max (Preview) 19% none Apr 2026 Epoch AI run
53 GLM-5.2 19% low Jun 2026 Epoch AI run
54 Gemini 3.5 Flash-Lite 19% low Jul 2026 Epoch AI run
55 Qwen3.5 397B-A17B 18% none Feb 2026 Epoch AI run
56 Kimi K2.6 18% Apr 2026 Epoch AI run
57 Qwen 3.6 Flash 18% 2026-04-16 Apr 2026 Epoch AI run
58 Claude Sonnet 4.5 17% 48k thinking Sep 2025 Epoch AI run
59 Qwen 3.5 Plus (hosted 397B-A17B) 17% none Feb 2026 Epoch AI run
60 Qwen3.5-122B-A10B 17% none Feb 2026 Epoch AI run
61 DeepSeek-V4-Pro 17% none Apr 2026 Epoch AI run
62 Qwen3.7-Plus 17% Jun 2026 Epoch AI run
63 Mistral Large 4 Preview 17% high Oct 2026 Epoch AI run
64 Claude Sonnet 4.6 16% low Feb 2026 Epoch AI run
65 Qwen3.7 Flash 15% Jul 2026 Epoch AI run
66 GPT-4 Jun 2023 12% Jun 2023 Epoch AI run
67 GPT-4o mini 12% Jul 2024 Epoch AI run
68 Qwen 3.6 Plus 12% 2026-04-02 Mar 2026 Epoch AI run
69 GPT-5.4 Mini 11% none Mar 2026 Epoch AI run
70 GPT-5 mini 10% minimal Aug 2025 Epoch AI run
71 Qwen3-235B-A22B-Thinking Jul 2025 9.00% Jul 2025 Epoch AI run
72 GPT-5 nano 9.00% medium Aug 2025 Epoch AI run
73 GPT-5.4 Nano 9.00% none Mar 2026 Epoch AI run
74 MiniMax-M3 8.00% none Jun 2026 Epoch AI run
75 GLM-5.3-Flash 8.00% max Aug 2026 Epoch AI run
76 o3-mini 7.00% high Jan 2025 Epoch AI run
77 GPT-4.1 mini 7.00% Apr 2025 Epoch AI run
78 Qwen3.6 27B 7.00% none Apr 2026 Epoch AI run
79 GPT-6 Luna 7.00% max Sep 2026 Epoch AI run
80 Inkling-Small 6.00% xhigh Jul 2026 Epoch AI run
81 o4-mini 5.00% high Apr 2025 Epoch AI run
82 Qwen3-Max 5.00% Sep 2025 Epoch AI run
83 GPT-3.5 Turbo Jan 2024 3.00% Jan 2024 Epoch AI run
84 gpt-oss-120b 2.00% medium Aug 2025 Epoch AI run

5 models have held number one in Epoch AI’s runs since its release in July 2026

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI Anthropic
Epoch AI runs Passed 85% in Nov 2026, by Claude Fable 5.2
Claude Fable 5.2 90%

Source: Epoch AI runs (CC BY 4.0)

Epoch AI’s run of Claude Fable 5.2 passed 85% in November 2026

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

Run by Epoch AI Best so far
0% 20% 40% 60% 80% 100% 2023 2024 2025 2026 2027 Near the ceiling (85%) Random guessing Claude Fable 5.2 GPT-6 Astra Claude Opus 5 GPT-5.4 o3 GPT-4 0% 20% 40% 60% 80% 100% 2023 2024 2025 2026 2027 Near the ceiling (85%) Random guessing Claude Fable 5.2 Claude Opus 5

Source: Epoch AI runs (CC BY 4.0)

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Games and long tasks

All 82 in Games and long tasks
  • ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6 Astra 99.95%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games. Gemini 2.5 Pro 3.90
  • MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks. Seed2.0 Pro 49%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress. Llama 4 Scout 7.00
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
  • Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%