Skip to main content
BenchLM
Data

Mystery Game Puzzles (Epoch AI) (Mystery Game Puzzles)

We show this table for reference; we do not rank on it.

Data verified

Epoch AI's Mystery Game Puzzles benchmark asks models for the single best next move in 100 positions from a puzzle variant of a well-known game whose identity Epoch keeps secret.

Benchmark score on Mystery Game Puzzles — November 13, 2026

We compile the Mystery Game Puzzles rows from benchmark-owner or independent runs. Claude Fable 5.2 leads the table at 90.0%, followed by GPT-6 Astra (84.0%) and GPT-6.1 Sol (80.0%). We do not use these results to rank models overall.

79 modelsReasoningCurrentDisplay onlyUpdated November 13, 2026

Benchmark score table (79 models)

Benchmark score results for Mystery Game Puzzles
RankModel / configurationScoreParameters (B)Open / closed
1Claude Fable 5.2AnthropicRun by Mystery Game Puzzles
90.0%
Not reportedClosed
2GPT-6 AstraOpenAIRun by Mystery Game Puzzles
84.0%
Not reportedClosed
3GPT-6.1 SolOpenAIRun by Mystery Game Puzzles
80.0%
Not reportedClosed
4GPT-6.1 TerraOpenAIRun by Mystery Game Puzzles
74.0%
Not reportedClosed
5Gemini 4 CarbonGoogleRun by Mystery Game Puzzles
73.0%
Not reportedClosed
6Claude Opus 5.5AnthropicRun by Mystery Game Puzzles
71.0%
Not reportedClosed
7Claude Sonnet 5.5AnthropicRun by Mystery Game Puzzles
65.0%
Not reportedClosed
8Gemini 4 ArgonGoogleRun by Mystery Game Puzzles
64.0%
Not reportedClosed
9Claude Opus 5AnthropicRun by Mystery Game Puzzles
59.0%
Not reportedClosed
10Claude Fable 5.1AnthropicRun by Mystery Game Puzzles
58.0%
Not reportedClosed
11GPT-5.6 SolOpenAIRun by Mystery Game Puzzles
58.0%
Not reportedClosed
12GPT-5.5OpenAIRun by Mystery Game Puzzles
56.0%
Not reportedClosed
13GPT-6 SolOpenAIRun by Mystery Game Puzzles
56.0%
Not reportedClosed
14Claude Fable 5AnthropicRun by Mystery Game Puzzles
52.0%
Not reportedClosed
15Muse WonderMetaRun by Mystery Game Puzzles
51.0%
Not reportedClosed
16DeepSeek V4.1 Flash 1017DeepSeekRun by Mystery Game Puzzles
49.0%
Not reportedOpen
17Gemini 3.8 FlashGoogleRun by Mystery Game Puzzles
47.0%
Not reportedClosed
18DeepSeek V4 Pro 0813DeepSeekRun by Mystery Game Puzzles
43.0%
Not reportedOpen
19Kimi K3.1 ProMoonshot AIRun by Mystery Game Puzzles
42.0%
Not reportedOpen
20Qwen3.8 MaxAlibabaRun by Mystery Game Puzzles
38.0%
Not reportedOpen
21GPT-5.4OpenAIRun by Mystery Game Puzzles
37.0%
Not reportedClosed
22Gemini 3.7 FlashGoogleRun by Mystery Game Puzzles
37.0%
Not reportedClosed
23Claude Opus 4.8AnthropicRun by Mystery Game Puzzles
36.0%
Not reportedClosed
24Claude Sonnet 5AnthropicRun by Mystery Game Puzzles
35.0%
Not reportedClosed
25GPT-5.6 TerraOpenAIRun by Mystery Game Puzzles
35.0%
Not reportedClosed
26Gemini 3.1 ProGoogleRun by Mystery Game Puzzles
34.0%
Not reportedClosed
27Grok 4.6xAIRun by Mystery Game Puzzles
34.0%
Not reportedClosed
28DeepSeek V4 Flash 0731DeepSeekRun by Mystery Game Puzzles
34.0%
Not reportedOpen
29GLM-5.3Z.AIRun by Mystery Game Puzzles
33.0%
Not reportedOpen
30Qwen3.7 MaxAlibabaRun by Mystery Game Puzzles
32.0%
Not reportedClosed
31Gemini 3.5 FlashGoogleRun by Mystery Game Puzzles
32.0%
Not reportedClosed
32Claude Haiku 5.5AnthropicRun by Mystery Game Puzzles
30.0%
Not reportedClosed
33Gemini 3.6 FlashGoogleRun by Mystery Game Puzzles
30.0%
Not reportedClosed
34o3OpenAIRun by Mystery Game Puzzles
29.0%
Not reportedClosed
35Grok 4.7xAIRun by Mystery Game Puzzles
29.0%
Not reportedClosed
36Claude Opus 4.7AnthropicRun by Mystery Game Puzzles
28.0%
Not reportedClosed
37Kimi K3Moonshot AIRun by Mystery Game Puzzles
26.0%
Not reportedOpen
38Claude Opus 4.6AnthropicRun by Mystery Game Puzzles
25.0%
Not reportedClosed
39Muse Spark 1.3MetaRun by Mystery Game Puzzles
25.0%
Not reportedClosed
40GPT-5.2OpenAIRun by Mystery Game Puzzles
23.0%
Not reportedClosed
41GPT-5 (high)OpenAIRun by Mystery Game Puzzles
23.0%
Not reportedClosed
42Claude Opus 4.5AnthropicRun by Mystery Game Puzzles
22.0%
Not reportedClosed
43Qwen3.6-35B-A3BAlibabaRun by Mystery Game Puzzles
22.0%
Not reportedOpen
44GPT-5.6 LunaOpenAIRun by Mystery Game Puzzles
21.0%
Not reportedClosed
45Nemotron 3 UltraNVIDIARun by Mystery Game Puzzles
20.0%
Not reportedOpen
46Gemini 3 FlashGoogleRun by Mystery Game Puzzles
20.0%
Not reportedClosed
47Qwen3.5 FlashAlibabaRun by Mystery Game Puzzles
20.0%
Not reportedClosed
48Gemini 3.5 Flash-LiteGoogleRun by Mystery Game Puzzles
19.0%
Not reportedClosed
49Qwen 3.6 Max (preview)AlibabaRun by Mystery Game Puzzles
19.0%
Not reportedClosed
50Kimi K2.6Moonshot AIRun by Mystery Game Puzzles
18.0%
Not reportedOpen
51Qwen3.5 397BAlibabaRun by Mystery Game Puzzles
18.0%
Not reportedOpen
52Qwen3.7 PlusAlibabaRun by Mystery Game Puzzles
17.0%
Not reportedClosed
53Qwen3.5-122B-A10BAlibabaRun by Mystery Game Puzzles
17.0%
Not reportedOpen
54Mistral Large 4MistralRun by Mystery Game Puzzles
17.0%
Not reportedOpen
55Claude Sonnet 4.5AnthropicRun by Mystery Game Puzzles
17.0%
Not reportedClosed
56Qwen3.5 PlusAlibabaRun by Mystery Game Puzzles
17.0%
Not reportedClosed
57DeepSeek V4 ProDeepSeekRun by Mystery Game Puzzles
17.0%
Not reportedOpen
58GPT-5.1OpenAIRun by Mystery Game Puzzles
16.0%
Not reportedClosed
59Claude Sonnet 4.6AnthropicRun by Mystery Game Puzzles
16.0%
Not reportedClosed
60GLM-5.2Z.AIRun by Mystery Game Puzzles
15.0%
Not reportedOpen
61GPT-5 (medium)OpenAIRun by Mystery Game Puzzles
15.0%
Not reportedClosed
62Qwen3.7 FlashAlibabaRun by Mystery Game Puzzles
15.0%
Not reportedClosed
63Qwen3.6 PlusAlibabaRun by Mystery Game Puzzles
12.0%
Not reportedClosed
64GPT-4o miniOpenAIRun by Mystery Game Puzzles
12.0%
Not reportedClosed
65Qwen3 235B 2507 (Reasoning)AlibabaRun by Mystery Game Puzzles
9.0%
Not reportedOpen
66MiniMax M3MiniMaxRun by Mystery Game Puzzles
8.0%
Not reportedOpen
67GLM-5.3-FlashZ.AIRun by Mystery Game Puzzles
8.0%
Not reportedOpen
68GPT-5 nanoOpenAIRun by Mystery Game Puzzles
8.0%
Not reportedClosed
69GPT-5.4 miniOpenAIRun by Mystery Game Puzzles
7.0%
Not reportedClosed
70Qwen3.6-27BAlibabaRun by Mystery Game Puzzles
7.0%
Not reportedOpen
71GPT-6 LunaOpenAIRun by Mystery Game Puzzles
7.0%
Not reportedClosed
72o3-miniOpenAIRun by Mystery Game Puzzles
7.0%
Not reportedClosed
73GPT-4.1 miniOpenAIRun by Mystery Game Puzzles
7.0%
Not reportedClosed
74Inkling-SmallThinking Machines LabRun by Mystery Game Puzzles
6.0%
Not reportedOpen
75GPT-5.4 nanoOpenAIRun by Mystery Game Puzzles
5.0%
Not reportedClosed
76o4-mini (high)OpenAIRun by Mystery Game Puzzles
5.0%
Not reportedClosed
77Qwen3 MaxAlibabaRun by Mystery Game Puzzles
5.0%
Not reportedClosed
78GPT-5 miniOpenAIRun by Mystery Game Puzzles
5.0%
Not reportedClosed
79GPT-OSS 120BOpenAIRun by Mystery Game Puzzles
0.0%
Not reportedOpen
  • Lab-reported
  • Benchmark-run
  • Independent
What the labels mean

How to read this leaderboard

Higher is better. With 100 positions, one answer is one point, so gaps of a few points are within noise. Claude Fable 5.2 leads at 90%.

Operator receipt: 79 sourced rows are currently displayable on this page; the leading published row is Claude Fable 5.2 at 90.0%.

Honest limit: Because the game, prompt and transcripts are secret, outside readers cannot inspect the task or check individual answers. Like Chess Puzzles it measures a narrow puzzle skill. BenchLM shows it as display-only; it does not enter overall or category rankings.

Among the reported Mystery Game Puzzles rows, Claude Fable 5.2 is first at 90.0%. The third row is 10.0 points behind. The broader top-10 range is 32.0 points, so the table still separates the published systems.

79 models have been evaluated on Mystery Game Puzzles. The benchmark falls in the Reasoning category. We keep these results separate from the weighted global scoring system, so they remain source-specific evidence. Mystery Game Puzzles is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Mystery Game Puzzles

Year

2026

Tasks

100 single-best-move positions from an undisclosed game

Format

Exact-match move answer from text input, minimal agent scaffold

Difficulty

Unfamiliar game positions, single best move

The setup mirrors Epoch's Chess Puzzles: positions come from random games rolled forward to mid-game and filtered to states with one best move. Models receive the state as text in a minimal agent scaffold and submit a move through a tool. Answers are normalized to the game's standard notation and must match the key; an illegal move is wrong and an answer with no valid move gets no credit. Epoch does not publish the prompt, examples or transcripts, to keep the game hidden and make benchmark-specific preparation harder.

Freshness and provenance

Version

Mystery Game Puzzles 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

Why is the game secret?

Epoch keeps the game's identity undisclosed so labs cannot prepare for the benchmark specifically. It is otherwise built like Epoch's Chess Puzzles.

Which model leads Mystery Game Puzzles?

Claude Fable 5.2 at max effort leads with 90%, ahead of GPT-6 Astra (84%), GPT-6.1 Sol (80%) and GPT-6.1 Terra (74%).

Last updated: November 13, 2026 · BenchLM version Mystery Game Puzzles 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.