Furniture Assembly
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 von 39 Ergebnissen; die Tabelle führt alle auf.
| # | Modell | Organisation | Ergebnis | Nachweis |
|---|---|---|---|---|
| 1 | Claude Fable 5.2 (max) API access | Anthropic | 90 % ± 3,9 % | – |
| 2 | Claude Opus 5.5 (max) API access | Anthropic | 83,3 % ± 4,7 % | – |
| 3 | GPT-6 Astra (max) API access | OpenAI | 80 % ± 4,9 % | – |
| 4 | GPT-6.1 Sol (max) API access | OpenAI | 80 % ± 4,9 % | – |
| 5 | GPT-6.1 Terra (max) API access | OpenAI | 79,2 % ± 5,1 % | – |
| 6 | Gemini 4 Carbon (high) API access | Google DeepMind | 77,5 % ± 5,4 % | – |
| 7 | Claude Sonnet 5.5 (max) API access | Anthropic | 75 % ± 5,5 % | – |
| 8 | Gemini 4 Argon (high) API access | Google DeepMind | 72,5 % ± 5,6 % | – |
| 9 | Claude Fable 5.1 (max) API access | Anthropic | 70 % ± 5,7 % | – |
| 10 | Claude Opus 5 (max) API access | Anthropic | 60,8 % ± 6,2 % | – |
| 11 | GPT-6 Sol (max) API access | OpenAI | 58,3 % ± 6,4 % | – |
| 12 | GPT-5.6 Sol (max) API access | OpenAI | 56,7 % ± 6,3 % | – |
| 13 | Kimi K3.1 Pro (Together, max) Open weights (restricted use) | Moonshot | 55,8 % ± 6,4 % | – |
| 14 | GPT-5.6 Terra (max) API access | OpenAI | 54,2 % ± 6,4 % | – |
| 15 | Muse Wonder (max) API access | Meta AI | 54,2 % ± 6,4 % | – |
| 16 | Claude Haiku 5.5 (max) API access | Anthropic | 47,5 % ± 6,4 % | – |
| 17 | GPT-5.5 (xhigh) API access | OpenAI | 44,2 % ± 6,4 % | – |
| 18 | GPT-6 Luna (max) API access | OpenAI | 44,2 % ± 6,4 % | – |
| 19 | Claude Opus 4.8 API access | Anthropic | 42,5 % ± 6,4 % | – |
| 20 | GPT-5.6 Luna (max) API access | OpenAI | 42,5 % ± 6,4 % | – |
| 21 | Grok 4.6 (xhigh) API access | xAI | 40 % ± 6,3 % | – |
| 22 | GPT-5.2 (xhigh) API access | OpenAI | 38,3 % ± 6,3 % | – |
| 23 | GPT-5.4 (xhigh) API access | OpenAI | 37,5 % ± 6,2 % | – |
| 24 | Claude Fable 5 (max) API access | Anthropic | 35,8 % ± 6,2 % | – |
| 25 | DeepSeek V4.1 Flash (max) Open weights (unrestricted) | DeepSeek | 34,2 % ± 6 % | – |
| 26 | Kimi K3 (max) Open weights (non-commercial) | Moonshot | 34,2 % ± 6,1 % | – |
| 27 | Claude Opus 4.7 (max) API access | Anthropic | 33,3 % ± 6,1 % | – |
| 28 | Gemini 3.8 Flash (high) API access | Google DeepMind | 31,7 % ± 5,9 % | – |
| 29 | Mistral Large 4 (high) Open weights (restricted use) | Mistral AI | 30,8 % ± 6 % | – |
| 30 | Claude Opus 4.5 (64k thinking) API access | Anthropic | 28,3 % ± 5,9 % | – |
| 31 | Claude Opus 4.6 (max) API access | Anthropic | 28,3 % ± 5,9 % | – |
| 32 | Gemini 3.1 Pro Preview API access | Google DeepMind | 26,7 % ± 5,6 % | – |
| 33 | Gemini 3.7 Flash (high) API access | Google DeepMind | 26,7 % ± 5,6 % | – |
| 34 | Gemini 3.6 Flash API access | Google DeepMind | 23,3 % ± 5,4 % | – |
| 35 | Grok 4.5 (high) API access | xAI | 22,5 % ± 5,4 % | – |
| 36 | Kimi K2.6 Open weights (unrestricted) | Moonshot | 21,7 % ± 5,2 % | – |
| 37 | Grok 4.7 (xhigh) API access | xAI | 20,8 % ± 5,2 % | – |
| 38 | Mistral Large 4 Preview (high) Open weights (restricted use) | Mistral AI | 20,8 % ± 5,2 % | – |
| 39 | Qwen3.8 Max (0902) (xhigh) API access | Alibaba | 20 % ± 5,2 % | – |
