llmrun
Reasoning

Mystery Game Puzzles Leaderboard

Mystery Game Puzzles is an Epoch AI evaluation of multi-step deductive reasoning: the model works through a mystery or logic-game scenario to arrive at the correct solution. It probes sustained, structured inference rather than single-step question answering.

Source: epoch17 open models ranked+105 proprietaryData through Sep 2026

Open modelsAll models

Open models ranked on Mystery Game Puzzles

# shows rank among open models / rank overall (including proprietary).

#ModelScore
1 / 9DeepSeek V4 Pro 0813 · 1650.5B
43.0%
2 / 17DeepSeek V4 Flash 0731 · 304.2B
34.0%
3 / 20GLM 5.3 · 753.3B
33.0%
4 / 35Kimi K3 · 2779.9B
26.0%
5 / 47Qwen3.6 35B A3B · 36.0B
22.0%
6 / 52NVIDIA Nemotron 3 Ultra 550B A55B BF16 · 560.5B
20.0%
7 / 55GLM 5.2 · 753.3B
19.0%
8 / 60Kimi K2.6 · 1026.9B
18.0%
9 / 61Qwen3.5 397B A17B · 403.4B
18.0%
10 / 64DeepSeek V4 Pro · 1598.8B
17.0%
11 / 67Qwen3.5 122B A10B · 125.1B
17.0%
12 / 102Qwen3 235B A22B Thinking 2507 · 235.1B
9.0%
13 / 103GLM 5.3 Flash · 321.3B
8.0%
14 / 106MiniMax M3 · 427.0B
8.0%
15 / 111Qwen3.6 27B · 27.8B
7.0%
16 / 113Inkling Small · 266.0B
6.0%
17 / 122GPT OSS 120B · 116.8B
2.0%

Score vs model size

Which models give the most quality for their size — the ones worth running locally.

100B1Tmodel size (log scale) →43.0%2.0%GLM 5.3 · 753B · 33.0%Kimi K3 · 2.8T · 26.0%NVIDIA Nemotron 3 Ultra 550B A55B BF16 · 561B · 20.0%GLM 5.2 · 753B · 19.0%Kimi K2.6 · 1T · 18.0%Qwen3.5 397B A17B · 403B · 18.0%DeepSeek V4 Pro · 1.6T · 17.0%Qwen3.5 122B A10B · 125B · 17.0%Qwen3 235B A22B Thinking 2507 · 235B · 9.0%MiniMax M3 · 427B · 8.0%GLM 5.3 Flash · 321B · 8.0%Inkling Small · 266B · 6.0%GPT OSS 120B · 117B · 2.0%Qwen3.6 27B · 28B · 7.0%Qwen3.6 27BQwen3.6 35B A3B · 36B · 22.0%Qwen3.6 35B A3BDeepSeek V4 Flash 0731 · 304B · 34.0%DeepSeek V4 Flash 0731DeepSeek V4 Pro 0813 · 1.7T · 43.0%DeepSeek V4 Pro 0813
Each dot is a model. Up = higher score, left = smaller (easier to run locally). The dashed line marks the efficiency frontier — the best score you can get at each size or smaller.
  • Qwen3.6 27B, 28B, score 7.0% — on the efficiency frontier (best score at its size or smaller).
  • Qwen3.6 35B A3B, 36B, score 22.0% — on the efficiency frontier (best score at its size or smaller).
  • DeepSeek V4 Flash 0731, 304B, score 34.0% — on the efficiency frontier (best score at its size or smaller).
  • DeepSeek V4 Pro 0813, 1.7T, score 43.0% — on the efficiency frontier (best score at its size or smaller).

Mystery Game Puzzles: frequently asked questions

What is the best open LLM on Mystery Game Puzzles?
DeepSeek V4 Pro 0813 is the top open model on Mystery Game Puzzles, scoring 43.0%. Among all models tested — including proprietary ones — it ranks #9. The top model overall is GPT 6 Astra Max (OpenAI) at 84.0%.
What's the best Mystery Game Puzzles model you can run on a 24 GB GPU?
Qwen3.6 35B A3B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (about 20 GB), scoring 22.0% on Mystery Game Puzzles.
Can open models match proprietary models on Mystery Game Puzzles?
Not quite on Mystery Game Puzzles: the strongest proprietary model (GPT 6 Astra Max) scores 84.0%, ahead of the best open model (DeepSeek V4 Pro 0813) at 43.0% — but you can run the open one yourself.

Scores aggregated from epoch. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.

Mystery Game Puzzles Leaderboard — LLM Scores | llmrun