Mystery Game Puzzles
Das Ergebnis jedes Modells, so, wie der Benchmarking Hub von Epoch AI es veröffentlicht, von Epoch gemessen oder aus der genannten Quelle gesammelt. Facthory hat sie nicht selbst gemessen.
Daten: Epoch AI, „Capabilities & benchmarking“, epoch.ai/benchmarks, lizenziert unter CC BY 4.0. Gelesen am 16.11.2026.
Das Diagramm zeigt die 25 besten, die Tabelle die 100 besten von 146 Ergebnissen.
| # | Modell | Organisation | Ergebnis | Nachweis |
|---|---|---|---|---|
| 1 | Claude Fable 5.2 (max) API access | Anthropic | 90 % ± 3 % | – |
| 2 | GPT-6 Astra (max) API access | OpenAI | 84 % ± 3,7 % | – |
| 3 | GPT-6.1 Sol (max) API access | OpenAI | 80 % ± 4 % | – |
| 4 | GPT-6.1 Terra (max) API access | OpenAI | 74 % ± 4,4 % | – |
| 5 | Gemini 4 Carbon (high) API access | Google DeepMind | 73 % ± 4,5 % | – |
| 6 | Claude Opus 5.5 (max) API access | Anthropic | 71 % ± 4,6 % | – |
| 7 | Claude Sonnet 5.5 (max) API access | Anthropic | 65 % ± 4,8 % | – |
| 8 | Gemini 4 Argon (high) API access | Google DeepMind | 64 % ± 4,8 % | – |
| 9 | Claude Opus 5 (max) API access | Anthropic | 59 % ± 4,9 % | – |
| 10 | Claude Fable 5.1 (max) API access | Anthropic | 58 % ± 5 % | – |
| 11 | GPT-5.6 Sol (max) API access | OpenAI | 58 % ± 5 % | – |
| 12 | GPT-5.5 (xhigh) API access | OpenAI | 56 % ± 5 % | – |
| 13 | GPT-6 Sol (max) API access | OpenAI | 56 % ± 5 % | – |
| 14 | Claude Fable 5 (max) API access | Anthropic | 52 % ± 5 % | – |
| 15 | GPT-5.5 (high) API access | OpenAI | 52 % ± 5 % | – |
| 16 | Muse Wonder (max) API access | Meta AI | 51 % ± 5 % | – |
| 17 | DeepSeek V4.1 Flash 1017 (max) Open weights (unrestricted) | DeepSeek | 49 % ± 5 % | – |
| 18 | Gemini 3.8 Flash (high) API access | Google DeepMind | 47 % ± 5 % | – |
| 19 | DeepSeek V4 Pro 0813 (max) Open weights (unrestricted) | DeepSeek | 43 % ± 5 % | – |
| 20 | DeepSeek V4.1 Flash (max) Open weights (unrestricted) | DeepSeek | 43 % ± 5 % | – |
| 21 | Kimi K3.1 Pro (Together, max) Open weights (restricted use) | Moonshot | 42 % ± 5 % | – |
| 22 | Qwen3.8 Max (xhigh) API access | Alibaba | 38 % ± 4,9 % | – |
| 23 | Claude Opus 5 API access | Anthropic | 37 % ± 4,9 % | – |
| 24 | Gemini 3.7 Flash (high) API access | Google DeepMind | 37 % ± 4,9 % | – |
| 25 | GPT-5.4 (xhigh) API access | OpenAI | 37 % ± 4,9 % | – |
| 26 | Claude Opus 4.8 API access | Anthropic | 36 % ± 4,8 % | – |
| 27 | Claude Sonnet 5 (max) API access | Anthropic | 35 % ± 4,8 % | – |
| 28 | GPT-5.6 Terra (max) API access | OpenAI | 35 % ± 4,8 % | – |
| 29 | DeepSeek V4 Flash 0731 (max) Open weights (unrestricted) | DeepSeek | 34 % ± 4,8 % | – |
| 30 | Gemini 3.1 Pro Preview API access | Google DeepMind | 34 % ± 4,8 % | – |
| 31 | Grok 4.6 (xhigh) API access | xAI | 34 % ± 4,8 % | – |
| 32 | GLM-5.3 (max) Open weights (restricted use) | Z.ai (Zhipu AI) | 33 % ± 4,7 % | – |
| 33 | GPT-5.6 Sol (none) API access | OpenAI | 33 % ± 4,7 % | – |
| 34 | Gemini 3.1 Pro Preview API access | Google DeepMind | 32 % ± 4,7 % | – |
| 35 | Gemini 3.5 Flash API access | Google DeepMind | 32 % ± 4,7 % | – |
| 36 | Qwen3.7 Max API access | Alibaba | 32 % ± 4,7 % | – |
| 37 | Claude Opus 4.8 API access | Anthropic | 31 % ± 4,6 % | – |
| 38 | Claude Haiku 5.5 (max) API access | Anthropic | 30 % ± 4,6 % | – |
| 39 | Gemini 3.6 Flash API access | Google DeepMind | 30 % ± 4,6 % | – |
| 40 | Gemini 3.1 Pro Preview API access | Google DeepMind | 29 % ± 4,6 % | – |
| 41 | Grok 4.7 (xhigh) API access | xAI | 29 % ± 4,6 % | – |
| 42 | o3 (high) API access | OpenAI | 29 % ± 4,6 % | – |
| 43 | Claude Opus 4.7 (max) API access | Anthropic | 28 % ± 4,5 % | – |
| 44 | Gemini 3.5 Flash API access | Google DeepMind | 28 % ± 4,5 % | – |
| 45 | GPT-5.4 (medium) API access | OpenAI | 28 % ± 4,5 % | – |
| 46 | GPT-5.5 (low) API access | OpenAI | 28 % ± 4,5 % | – |
| 47 | Gemini 3 Flash Preview API access | Google DeepMind | 26 % ± 4,4 % | – |
| 48 | GPT-5.6 Sol (low) API access | OpenAI | 26 % ± 4,4 % | – |
| 49 | Kimi K3 (max) Open weights (non-commercial) | Moonshot | 26 % ± 4,4 % | – |
| 50 | qwen3.7-max_none API access | Alibaba | 26 % ± 4,4 % | – |
| 51 | Claude Opus 4.6 (max) API access | Anthropic | 25 % ± 4,4 % | – |
| 52 | Gemini 3 Flash Preview API access | Google DeepMind | 25 % ± 4,4 % | – |
| 53 | Gemini 3.6 Flash API access | Google DeepMind | 25 % ± 4,4 % | – |
| 54 | Muse Spark 1.3 (max) API access | Meta AI | 25 % ± 4,5 % | – |
| 55 | Gemini 3.6 Flash API access | Google DeepMind | 23 % ± 4,2 % | – |
| 56 | GPT-5 (high) API access | OpenAI | 23 % ± 4,2 % | – |
| 57 | GPT-5.2 (high) API access | OpenAI | 23 % ± 4,2 % | – |
| 58 | o3 (medium) API access | OpenAI | 23 % ± 4,2 % | – |
| 59 | Claude Opus 4.5 (48k thinking) API access | Anthropic | 22 % ± 4,2 % | – |
| 60 | GPT-5.2 (medium) API access | OpenAI | 22 % ± 4,2 % | – |
| 61 | Qwen3.6 35B-A3B Open weights (unrestricted) | Alibaba | 22 % ± 4,2 % | – |
| 62 | claude-opus-4-1-20250805_24K API access | Anthropic | 21 % ± 4,1 % | – |
| 63 | GPT-5.6 Luna (max) API access | OpenAI | 21 % ± 4,1 % | – |
| 64 | Gemini 3 Flash Preview API access | Google DeepMind | 20 % ± 4 % | – |
| 65 | GPT-5.6 Luna (none) API access | OpenAI | 20 % ± 4 % | – |
| 66 | Mistral Large 4 (high) Open weights (restricted use) | Mistral AI | 20 % ± 4 % | – |
| 67 | nemotron-3-ultra Open weights (unrestricted) | Nvidia | 20 % ± 4 % | – |
| 68 | Qwen3.5 Flash | Alibaba | 20 % ± 4 % | – |
| 69 | Gemini 3.5 Flash-Lite API access | Google DeepMind | 19 % ± 3,9 % | – |
| 70 | GLM-5.2 Open weights (unrestricted) | Z.ai (Zhipu AI) | 19 % ± 3,9 % | – |
| 71 | GPT-5.1 (low) API access | OpenAI | 19 % ± 3,9 % | – |
| 72 | GPT-5.6 Terra (low) API access | OpenAI | 19 % ± 3,9 % | – |
| 73 | Qwen 3.6 Max (Preview) | Alibaba | 19 % ± 3,9 % | – |
| 74 | glm-5.2_minimal Open weights (unrestricted) | Z.ai (Zhipu AI) | 18 % ± 3,9 % | – |
| 75 | GLM-5.2 Open weights (unrestricted) | Z.ai (Zhipu AI) | 18 % ± 3,9 % | – |
| 76 | GPT-5.5 (no thinking) API access | OpenAI | 18 % ± 3,9 % | – |
| 77 | Kimi K2.6 Open weights (unrestricted) | Moonshot | 18 % ± 3,9 % | – |
| 78 | Qwen3.5 397B-A17B Open weights (unrestricted) | Alibaba | 18 % ± 3,9 % | – |
| 79 | Qwen 3.6 Flash (2026-04-16) | Alibaba | 18 % ± 3,9 % | – |
| 80 | Claude Sonnet 4.5 (48k thinking) API access | Anthropic | 17 % ± 3,8 % | – |
| 81 | DeepSeek v4 Pro (unknown thinking) Open weights (unrestricted) | DeepSeek | 17 % ± 3,8 % | – |
| 82 | GPT-5.4 (low) API access | OpenAI | 17 % ± 3,8 % | – |
| 83 | GPT-5.6 Luna (low) API access | OpenAI | 17 % ± 3,8 % | – |
| 84 | Mistral Large 4 Preview (high) Open weights (restricted use) | Mistral AI | 17 % ± 3,8 % | – |
| 85 | qwen3.5-122b-a10b_none Open weights (restricted use) | Alibaba | 17 % ± 3,8 % | – |
| 86 | Qwen3.5 Plus | Alibaba | 17 % ± 3,8 % | – |
| 87 | Qwen3.7 Plus API access | Alibaba | 17 % ± 3,8 % | – |
| 88 | Qwen3.7 Plus API access | Alibaba | 17 % ± 3,8 % | – |
| 89 | Claude Sonnet 4.6 (low) API access | Anthropic | 16 % ± 3,7 % | – |
| 90 | Claude Sonnet 5 (no thinking) API access | Anthropic | 16 % ± 3,7 % | – |
| 91 | GPT-5 (low) API access | OpenAI | 16 % ± 3,7 % | – |
| 92 | GPT-5.1 (medium) API access | OpenAI | 16 % ± 3,7 % | – |
| 93 | GPT-5.4 (none) API access | OpenAI | 16 % ± 3,7 % | – |
| 94 | Qwen3.5 Flash | Alibaba | 16 % ± 3,7 % | – |
| 95 | Qwen3.5 Plus | Alibaba | 16 % ± 3,7 % | – |
| 96 | Claude Opus 4.6 (no thinking) API access | Anthropic | 15 % ± 3,6 % | – |
| 97 | GLM-5.2 Open weights (unrestricted) | Z.ai (Zhipu AI) | 15 % ± 3,6 % | – |
| 98 | GPT-5 (medium) API access | OpenAI | 15 % ± 3,6 % | – |
| 99 | GPT-5.1 (no thinking) API access | OpenAI | 15 % ± 3,6 % | – |
| 100 | Qwen3.7 Flash API access | Alibaba | 15 % ± 3,6 % | – |
