Benchmark · KnowledgeCompetitive

Mystery Game Puzzles

Updated 2026-11-10
Models tested
67
Top score
89.0
Claude Fable 5.2
Median
21.8
min 0.9
Top-5 spread
σ 7.0
wide open

Best score over time · one chart, every benchmark

MYSTERY GAME PUZZLES63 MODELS · FRONTIER RUNNING MAX0255075100SCORE ↑Apr 25Sep 25Jan 26Jun 26Nov 26RELEASE DATE →benchgecko.ai/benchmark/mystery-game-puzzles · frontier
Frontier on Mystery Game Puzzles rose from 21.8 to 89.0 in 19 months · +67.2 points · latest leader Claude Fable 5.2 from Anthropic.
Pink dots = frontier records · 8 totalClick to open model page

67 models tested · sorted by score

#ModelScore
1Anthropic logoClaude Fable 5.289.0
2OpenAI logoGPT-6 Astra82.4
3OpenAI logoGPT-6.1 Sol78.0
4OpenAI logoGPT-6.1 Terra71.4
5Google DeepMind logoGemini 4 Carbon70.3
6Anthropic logoClaude Opus 5.568.0
7Anthropic logoClaude Sonnet 5.561.5
8Google DeepMind logoGemini 4 Argon60.3
9Anthropic logoClaude Opus 554.8
10Anthropic logoClaude Fable 5.153.7
11OpenAI logoGPT-5.6 Sol53.7
12OpenAI logoGPT-5.551.5
13OpenAI logoGPT-6 Sol51.5
14Anthropic logoClaude Fable 547.1
15DeepSeek logoDeepSeek V4.1 Flash 101743.8
16Google DeepMind logoGemini 3.8 Flash41.6
17DeepSeek logoDeepSeek V4.1 Flash37.2
18DeepSeek logoDeepSeek V4 Pro 081337.2
19moonshotai logoKimi K3.1 Pro36.1
20Alibaba Qwen logoQwen3.8 Max (0902)31.7
21Google DeepMind logoGemini 3.7 Flash30.6
22OpenAI logoGPT-5.430.6
23Anthropic logoClaude Opus 4.829.5
24OpenAI logoGPT-5.6 Terra28.4
25Anthropic logoClaude Sonnet 528.4
26Google DeepMind logoGemini 3.1 Pro Preview27.3
27DeepSeek logoDeepSeek V4 Flash 073127.3
28xAI logoGrok 4.627.3
29z-ai logoGLM 5.326.2
30Google DeepMind logoGemini 3.5 Flash25.1
31Alibaba Qwen logoQwen3.7 Max25.1
32Google DeepMind logoGemini 3.6 Flash22.9
33Anthropic logoClaude Haiku 5.522.9
34xAI logoGrok 4.721.8
35OpenAI logoo321.8
36Anthropic logoClaude Opus 4.720.7
37moonshotai logoKimi K318.5
38Google DeepMind logoGemini 3 Flash Preview18.5
39Anthropic logoClaude Opus 4.617.4
40OpenAI logoGPT-5.215.2
41OpenAI logoGPT-515.2
42Anthropic logoClaude Opus 4.514.1
43Alibaba Qwen logoQwen3.6 35B A3B14.1
44Anthropic logoClaude Opus 4.113.0
45OpenAI logoGPT-5.6 Luna13.0
46NVIDIA logoNemotron 3 Ultra11.9
47Alibaba Qwen logoQwen3.5-Flash11.9
48Mistral AI logoMistral Large 411.9
49z-ai logoGLM 5.210.8
50OpenAI logoGPT-5.110.8
51Google DeepMind logoGemini 3.5 Flash Lite10.8
52Alibaba Qwen logoQwen3.6 Max Preview10.8
53moonshotai logoKimi K2.69.7
54Alibaba Qwen logoQwen3.6 Flash9.7
55Alibaba Qwen logoQwen3.5 397B A17B9.7
56Anthropic logoClaude Sonnet 4.58.6
57Alibaba Qwen logoQwen3.7 Plus8.6
58DeepSeek logoDeepSeek V4 Pro8.6
59Alibaba Qwen logoQwen 3.5 Plus (hosted 397B-A17B)8.6
60Anthropic logoClaude Sonnet 4.67.5
61Alibaba Qwen logoQwen3.7 Flash6.4
62OpenAI logoGPT-4 Turbo (batch)3.1
63Alibaba Qwen logoQwen3.6 Plus3.1
64OpenAI logoGPT-4o-mini3.1
65OpenAI logoGPT-4 Turbo3.1
66OpenAI logoGPT-5.4 Mini2.0
67OpenAI logoGPT-5 Mini0.9

What it measures · how it is scored · sources · as of 2026-11-17

  • How it is scored · Scored as a percentage · higher is better · maximum 100
  • Released · 2026-07-30
  • Scores from · Epoch AI Benchmarking Hub (67 scores)

Same category · related evaluations