DTBench
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 241 Ergebnissen.
| # | Modell | Organisation | Ergebnis | Nachweis |
|---|---|---|---|---|
| 1 | Claude Fable 5.2 (unknown) API access | Anthropic | 99,2 % | – |
| 2 | Claude Fable 5.2 (xhigh) API access | Anthropic | 98,9 % | – |
| 3 | Claude Opus 5.5 (max) API access | Anthropic | 98,9 % | – |
| 4 | Claude Fable 5 (max) API access | Anthropic | 98,4 % | – |
| 5 | Claude Fable 5.2 (high) API access | Anthropic | 98,1 % | – |
| 6 | Claude Fable 5.2 (low) API access | Anthropic | 98,1 % | – |
| 7 | Claude Fable 5.2 (medium) API access | Anthropic | 98,1 % | – |
| 8 | Claude Opus 5 (high) API access | Anthropic | 97,9 % | – |
| 9 | Claude Opus 5 (xhigh) API access | Anthropic | 97,9 % | – |
| 10 | Claude Fable 5.1 (unknown) API access | Anthropic | 97,6 % | – |
| 11 | Claude Fable 5.1 (xhigh) API access | Anthropic | 97,6 % | – |
| 12 | Claude Opus 5 (max) API access | Anthropic | 97,6 % | – |
| 13 | Gemini 4 Argon (low) API access | Google DeepMind | 97,6 % | – |
| 14 | Claude Fable 5 (xhigh) API access | Anthropic | 97,3 % | – |
| 15 | Claude Opus 5.5 (xhigh) API access | Anthropic | 97,3 % | – |
| 16 | Claude Sonnet 5.5 (max) API access | Anthropic | 97,3 % | – |
| 17 | Gemini 4 Argon (high) API access | Google DeepMind | 97,3 % | – |
| 18 | Gemini 4 Argon (medium) API access | Google DeepMind | 97,3 % | – |
| 19 | GPT-6 Astra (xhigh) API access | OpenAI | 97,3 % | – |
| 20 | GPT-6 Sol (max) API access | OpenAI | 97,3 % | – |
| 21 | Grok 4.6 (xhigh) API access | xAI | 97,3 % | – |
| 22 | Claude Fable 5.1 (medium) API access | Anthropic | 97,1 % | – |
| 23 | Gemini 3.1 Pro Preview API access | Google DeepMind | 97,1 % | – |
| 24 | GPT-6 Astra (medium) API access | OpenAI | 97,1 % | – |
| 25 | GPT-6 Astra (unknown thinking) API access | OpenAI | 97,1 % | – |
| 26 | Gemini 3.7 Flash (high) API access | Google DeepMind | 96,8 % | – |
| 27 | Claude Fable 5.1 (high) API access | Anthropic | 96,5 % | – |
| 28 | Claude Fable 5 (high) API access | Anthropic | 96,5 % | – |
| 29 | Claude Opus 5.5 (high) API access | Anthropic | 96,5 % | – |
| 30 | Claude Opus 5.5 (medium) API access | Anthropic | 96,5 % | – |
| 31 | GPT-6 Astra (high) API access | OpenAI | 96,5 % | – |
| 32 | GPT-6.1 Sol (max) API access | OpenAI | 96,5 % | – |
| 33 | Grok 4.5 (high) API access | xAI | 96,5 % | – |
| 34 | Muse Spark 1.3 (unknown) API access | Meta AI | 96,5 % | – |
| 35 | Claude Sonnet 5.5 (xhigh) API access | Anthropic | 96 % | – |
| 36 | GPT-5.5 Pro (xhigh) API access | OpenAI | 96 % | – |
| 37 | GPT-5.5 (xhigh) API access | OpenAI | 96 % | – |
| 38 | GPT-5.6 Sol (pro, max) API access | OpenAI | 96 % | – |
| 39 | Grok 4.7 (xhigh) API access | xAI | 96 % | – |
| 40 | Claude Fable 5 (medium) API access | Anthropic | 95,7 % | – |
| 41 | GPT-5.6 Sol (xhigh) API access | OpenAI | 95,7 % | – |
| 42 | Gemini 3.8 Flash (unknown) API access | Google DeepMind | 95,7 % | – |
| 43 | Claude Opus 5.5 (low) API access | Anthropic | 95,5 % | – |
| 44 | Gemini 3.1 Pro Preview API access | Google DeepMind | 95,5 % | – |
| 45 | Gemini 3.6 Flash API access | Google DeepMind | 95,5 % | – |
| 46 | GPT-5.6 Sol (high) API access | OpenAI | 95,5 % | – |
| 47 | GPT-5.6 Sol (max) API access | OpenAI | 95,5 % | – |
| 48 | GPT-5.6 Sol (medium) API access | OpenAI | 95,2 % | – |
| 49 | Muse Wonder (unknown) API access | Meta AI | 95,2 % | – |
| 50 | Claude Haiku 5.5 (max) API access | Anthropic | 95 % | – |
| 51 | Claude Opus 4.8 API access | Anthropic | 94,9 % | – |
| 52 | Claude Opus 5 (medium) API access | Anthropic | 94,9 % | – |
| 53 | Gemini 3.1 Pro Preview API access | Google DeepMind | 94,9 % | – |
| 54 | Gemini 3.8 Flash (low) API access | Google DeepMind | 94,9 % | – |
| 55 | Gemini 3.8 Flash (medium) API access | Google DeepMind | 94,9 % | – |
| 56 | GPT-6 Astra (low) API access | OpenAI | 94,9 % | – |
| 57 | Gemini 3.5 Flash API access | Google DeepMind | 94,7 % | – |
| 58 | Gemini 3.7 Flash (medium) API access | Google DeepMind | 94,7 % | – |
| 59 | Muse Spark 1.2 (xhigh) API access | Meta AI | 94,7 % | – |
| 60 | Claude Opus 4.7 (max) API access | Anthropic | 94,7 % | – |
| 61 | GPT-5.4 (xhigh) API access | OpenAI | 94,4 % | – |
| 62 | Muse Spark 1.1 (high) API access | Meta AI | 94,4 % | – |
| 63 | Claude Fable 5.1 (low) API access | Anthropic | 94,1 % | – |
| 64 | Claude Sonnet 5.5 (high) API access | Anthropic | 94,1 % | – |
| 65 | Claude Sonnet 5.5 (medium) API access | Anthropic | 93,9 % | – |
| 66 | DeepSeek V4 Pro 0813 (unknown) Open weights (unrestricted) | DeepSeek | 93,9 % | – |
| 67 | GPT-5.6 Sol (low) API access | OpenAI | 93,9 % | – |
| 68 | GLM-5.2 Open weights (unrestricted) | Z.ai (Zhipu AI) | 93,6 % | – |
| 69 | Claude Sonnet 5.5 (low) API access | Anthropic | 93,3 % | – |
| 70 | Gemini 3.7 Flash (low) API access | Google DeepMind | 93,3 % | – |
| 71 | GPT-5.6 Terra (max) API access | OpenAI | 93,3 % | – |
| 72 | Kimi K3.1 Pro (max) Open weights (restricted use) | Moonshot | 93,3 % | – |
| 73 | Claude Opus 5 (low) API access | Anthropic | 93,1 % | – |
| 74 | GPT-5.6 Terra (xhigh) API access | OpenAI | 92,5 % | – |
| 75 | Claude Sonnet 5 (max) API access | Anthropic | 92,5 % | – |
| 76 | Qwen3.7 Max API access | Alibaba | 92,3 % | – |
| 77 | Qwen3.8 Max (xhigh) API access | Alibaba | 92 % | – |
| 78 | Claude Fable 5 (low) API access | Anthropic | 91,5 % | – |
| 79 | Claude Opus 4.6 (max) API access | Anthropic | 91,2 % | – |
| 80 | GPT-5.6 Terra (high) API access | OpenAI | 91,2 % | – |
| 81 | Kimi K3 (max) Open weights (non-commercial) | Moonshot | 91,2 % | – |
| 82 | DeepSeek V4 Flash 0731 (max) Open weights (unrestricted) | DeepSeek | 90,9 % | – |
| 83 | GPT-5.2 (xhigh) API access | OpenAI | 90,9 % | – |
| 84 | Kimi K2.6 Open weights (unrestricted) | Moonshot | 90,9 % | – |
| 85 | DeepSeek v4 Pro (max) Open weights (unrestricted) | DeepSeek | 90,7 % | – |
| 86 | GPT-5 (high) API access | OpenAI | 90,7 % | – |
| 87 | grok-4.3_high Hosted access (no API) | xAI | 90,7 % | – |
| 88 | GPT-5.1 (high) API access | OpenAI | 90,1 % | – |
| 89 | GPT-5.6 Terra (low) API access | OpenAI | 90,1 % | – |
| 90 | GPT-5.6 Terra (medium) API access | OpenAI | 90,1 % | – |
| 91 | GPT-6 Luna (max) API access | OpenAI | 90,1 % | – |
| 92 | grok-4-20 API access | xAI | 90,1 % | – |
| 93 | nemotron-3-ultra Open weights (unrestricted) | Nvidia | 90,1 % | – |
| 94 | Claude Opus 4.5 (high) API access | Anthropic | 89,9 % | – |
| 95 | Claude Sonnet 4.6 (max) API access | Anthropic | 89,9 % | – |
| 96 | DeepSeek V4.1 Flash (unknown) Open weights (unrestricted) | DeepSeek | 89,9 % | – |
| 97 | Claude Sonnet 5 (xhigh) API access | Anthropic | 89,5 % | – |
| 98 | Gemini 3 Flash Preview API access | Google DeepMind | 89,1 % | – |
| 99 | GPT-5.6 Luna (xhigh) API access | OpenAI | 89,1 % | – |
| 100 | GPT-5.6 Luna (max) API access | OpenAI | 88,8 % | – |
