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: epoch20 open models ranked+116 proprietaryData through Oct 2026

Open modelsAll models

All models ranked on Mystery Game Puzzles

Proprietary / closed models are shown dimmed — you can't run them locally, but they show where the open field stands.

#ModelScoreVRAM @ Q4
1GPT 6 Astra Max · proprietary
84.0%
API only
2GPT 6.1 Sol Max · proprietary
80.0%
API only
3GPT 6.1 Terra Max · proprietary
74.0%
API only
4Claude Opus 5.5 Max · proprietary
71.0%
API only
5Claude Sonnet 5.5 Max · proprietary
65.0%
API only
6Gemini 4 Argon (high) · proprietary
64.0%
API only
7Claude Opus 5 Max · proprietary
59.0%
API only
8Claude Fable 5.1 Max · proprietary
58.0%
API only
9GPT 5.6 Sol Max · proprietary
58.0%
API only
10GPT 5.5 (xhigh) · proprietary
56.0%
API only
11GPT 6 Sol Max · proprietary
56.0%
API only
12Claude Fable 5 Max · proprietary
52.0%
API only
13GPT 5.5 (high) · proprietary
52.0%
API only
14Muse Wonder 1.0 Max · proprietary
51.0%
API only
15DeepSeek V4.1 Flash 1017 · 763.2B
49.0%
~458 GB
16Gemini 3.8 Flash (high) · proprietary
47.0%
API only
17DeepSeek V4 Pro 0813 · 1650.5B
43.0%
~991 GB
18DeepSeek V4.1 Flash · 763.2B
43.0%
~458 GB
19Kimi K3.1 Pro Together Max · proprietary
42.0%
API only
20Qwen3.8 Max (xhigh) · proprietary
38.0%
API only
21Claude Opus 5 · proprietary
37.0%
API only
22Gemini 3.7 Flash (high) · proprietary
37.0%
API only
23GPT 5.4 (Mar 05, 2026, xhigh) · proprietary
37.0%
API only
24Claude Opus 4.8 Max · proprietary
36.0%
API only
25Claude Sonnet 5 Max · proprietary
35.0%
API only
26GPT 5.6 Terra Max · proprietary
35.0%
API only
27DeepSeek V4 Flash 0731 · 304.2B
34.0%
~183 GB
28Gemini 3.1 Pro Preview (high) · proprietary
34.0%
API only
29Grok 4.6 (xhigh) · proprietary
34.0%
API only
30GLM 5.3 · 753.3B
33.0%
~456 GB
31GPT 5.6 Sol None · proprietary
33.0%
API only
32Gemini 3.1 Pro Preview (medium) · proprietary
32.0%
API only
33Gemini 3.5 Flash (high) · proprietary
32.0%
API only
34Qwen3.7 Max · proprietary
32.0%
API only
35Claude Opus 4.8 (xhigh) · proprietary
31.0%
API only
36Claude Haiku 5.5 Max · proprietary
30.0%
API only
37Gemini 3.6 Flash (high) · proprietary
30.0%
API only
38Gemini 3.1 Pro Preview (low) · proprietary
29.0%
API only
39Grok 4.7 (xhigh) · proprietary
29.0%
API only
40O3 (Apr 16, 2025, high) · proprietary
29.0%
API only
41Claude Opus 4.7 Max · proprietary
28.0%
API only
42Gemini 3.5 Flash (low) · proprietary
28.0%
API only
43GPT 5.4 (Mar 05, 2026, medium) · proprietary
28.0%
API only
44GPT 5.5 (low) · proprietary
28.0%
API only
45Gemini 3 Flash Preview (low) · proprietary
26.0%
API only
46GPT 5.6 Sol (low) · proprietary
26.0%
API only
47Kimi K3 · 2779.9B
26.0%
~1674 GB
48Qwen3.7 Max None · proprietary
26.0%
API only
49Claude Opus 4.6 Max · proprietary
25.0%
API only
50Gemini 3 Flash Preview (minimal) · proprietary
25.0%
API only
51Gemini 3.6 Flash (minimal) · proprietary
25.0%
API only
52Muse Spark 1.3 Max · proprietary
25.0%
API only
53Gemini 3.6 Flash (low) · proprietary
23.0%
API only
54GPT 5 (Aug 07, 2025, high) · proprietary
23.0%
API only
55GPT 5.2 (Dec 11, 2025, high) · proprietary
23.0%
API only
56O3 (Apr 16, 2025, medium) · proprietary
23.0%
API only
57Claude Opus 4.5 (Nov 01, 2025, 48K) · proprietary
22.0%
API only
58GPT 5.2 (Dec 11, 2025, medium) · proprietary
22.0%
API only
59Qwen3.6 35B A3B · 36.0B
22.0%
~22 GB
60Claude Opus 4.1 (Aug 05, 2025, 24K) · proprietary
21.0%
API only
61GPT 5.6 Luna Max · proprietary
21.0%
API only
62Gemini 3 Flash Preview (high) · proprietary
20.0%
API only
63GPT 5.6 Luna None · proprietary
20.0%
API only
64Mistral Large 4 1T A52B · 1000B
20.0%
~604 GB
65NVIDIA Nemotron 3 Ultra 550B A55B BF16 · 560.5B
20.0%
~370 GB
66Qwen3.5 Flash · proprietary
20.0%
API only
67Gemini 3.5 Flash Lite (low) · proprietary
19.0%
API only
68GLM 5.2 · 753.3B
19.0%
~456 GB
69GPT 5.1 (Nov 13, 2025, low) · proprietary
19.0%
API only
70GPT 5.6 Terra (low) · proprietary
19.0%
API only
71Qwen3.6 Max Preview None · proprietary
19.0%
API only
72GPT 5.5 None · proprietary
18.0%
API only
73Kimi K2.6 · 1026.9B
18.0%
~620 GB
74Qwen3.5 397B A17B · 403.4B
18.0%
~242 GB
75Qwen3.6 Flash None · proprietary
18.0%
API only
76Claude Sonnet 4.5 (Sep 29, 2025, 48K) · proprietary
17.0%
API only
77DeepSeek V4 Pro · 1598.8B
17.0%
~960 GB
78GPT 5.4 (Mar 05, 2026, low) · proprietary
17.0%
API only
79GPT 5.6 Luna (low) · proprietary
17.0%
API only
80Qwen3.5 122B A10B · 125.1B
17.0%
~75 GB
81Qwen3.5 Plus None · proprietary
17.0%
API only
82Qwen3.7 Plus · proprietary
17.0%
API only
83Qwen3.7 Plus None · proprietary
17.0%
API only
84Claude Sonnet 4.6 (low) · proprietary
16.0%
API only
85Claude Sonnet 5 · proprietary
16.0%
API only
86GPT 5 (Aug 07, 2025, low) · proprietary
16.0%
API only
87GPT 5.1 (Nov 13, 2025, medium) · proprietary
16.0%
API only
88GPT 5.4 2026 03.05 None · proprietary
16.0%
API only
89Qwen3.5 Flash None · proprietary
16.0%
API only
90Qwen3.5 Plus · proprietary
16.0%
API only
91Claude Opus 4.6 · proprietary
15.0%
API only
92GPT 5 (Aug 07, 2025, medium) · proprietary
15.0%
API only
93GPT 5.1 2025 11.13 None · proprietary
15.0%
API only
94Qwen3.7 Flash · proprietary
15.0%
API only
95Muse Spark 1.3 (xhigh) · proprietary
14.1%
API only
96Claude Sonnet 4.6 · proprietary
14.0%
API only
97GPT 5 (Aug 07, 2025, minimal) · proprietary
14.0%
API only
98GPT 5.2 2025 12.11 None · proprietary
14.0%
API only
99GPT 5.6 Terra None · proprietary
14.0%
API only
100Qwen3.7 Flash None · proprietary
14.0%
API only
101Claude Opus 4.7 · proprietary
13.0%
API only
102Gemini 3.5 Flash Lite (minimal) · proprietary
12.0%
API only
103GPT 4 (Jun 13) · proprietary
12.0%
API only
104GPT 4o Mini (Jul 18, 2024) · proprietary
12.0%
API only
105GPT 5.6 Luna (medium) · proprietary
12.0%
API only
106Qwen3.6 Flash · proprietary
12.0%
API only
107Qwen3.6 Plus None · proprietary
12.0%
API only
108GPT 5.4 Mini 2026 03.17 None · proprietary
11.0%
API only
109GPT 5 Mini (Aug 07, 2025, minimal) · proprietary
10.0%
API only
110GPT 5.2 (Dec 11, 2025, low) · proprietary
10.0%
API only
111GPT 5.6 Terra (medium) · proprietary
10.0%
API only
112GPT 5 Nano (Aug 07, 2025, low) · proprietary
9.0%
API only
113GPT 5 Nano (Aug 07, 2025, medium) · proprietary
9.0%
API only
114GPT 5.4 Nano 2026 03.17 None · proprietary
9.0%
API only
115Qwen3 235B A22B Thinking 2507 · 235.1B
9.0%
~142 GB
116GLM 5.3 Flash · 321.3B
8.0%
~195 GB
117GPT 5 Nano (Aug 07, 2025, high) · proprietary
8.0%
API only
118GPT 5.4 Mini (Mar 17, 2026, low) · proprietary
8.0%
API only
119MiniMax M3 · 427.0B
8.0%
~257 GB
120Claude Opus 4.6 (low) · proprietary
7.0%
API only
121GPT 4.1 Mini (Apr 14, 2025) · proprietary
7.0%
API only
122GPT 5.4 Mini (Mar 17, 2026, medium) · proprietary
7.0%
API only
123GPT 6 Luna Max · proprietary
7.0%
API only
124O3 Mini (Jan 31, 2025, high) · proprietary
7.0%
API only
125Qwen3.6 27B · 27.8B
7.0%
~17 GB
126GPT 5.4 Nano (Mar 17, 2026, medium) · proprietary
6.0%
API only
127Inkling Small · 266.0B
6.0%
~160 GB
128GPT 5 Mini (Aug 07, 2025, high) · proprietary
5.0%
API only
129GPT 5 Nano (Aug 07, 2025, minimal) · proprietary
5.0%
API only
130GPT 5.4 Nano (Mar 17, 2026, high) · proprietary
5.0%
API only
131O4 Mini (Apr 16, 2025, high) · proprietary
5.0%
API only
132Qwen3 Max (Sep 23, 2025) · proprietary
5.0%
API only
133GPT 5 Mini (Aug 07, 2025, medium) · proprietary
4.0%
API only
134GPT 3.5 Turbo (Jan 25) · proprietary
3.0%
API only
135GPT 5.4 Nano (Mar 17, 2026, low) · proprietary
3.0%
API only
136GPT OSS 120B · 116.8B
2.0%
~71 GB

Score vs model size

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

100B1Tmodel size (log scale) →49.0%2.0%DeepSeek V4.1 Flash · 763B · 43.0%DeepSeek V4 Pro 0813 · 1.7T · 43.0%GLM 5.3 · 753B · 33.0%Kimi K3 · 2.8T · 26.0%Mistral Large 4 1T A52B · 1T · 20.0%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 V4.1 Flash 1017 · 763B · 49.0%DeepSeek V4.1 Flash 1…
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 V4.1 Flash 1017, 763B, score 49.0% — 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 V4.1 Flash 1017 is the top open model on Mystery Game Puzzles, scoring 49.0%. Among all models tested — including proprietary ones — it ranks #15. 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 (~22 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 V4.1 Flash 1017) at 49.0% — 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.