Terminal Bench
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 59 Ergebnissen; die Tabelle führt alle auf.
| # | Modell | Organisation | Ergebnis | Nachweis |
|---|---|---|---|---|
| 1 | GPT-5.5 (unknown thinking) API access | OpenAI | 84,7 % | – |
| 2 | GPT-5.4 (unknown thinking) API access | OpenAI | 81,8 % | – |
| 3 | Claude Opus 4.7 (unknown) API access | Anthropic | 80,2 % | – |
| 4 | Gemini 3.1 Pro Preview API access | Google DeepMind | 80,2 % | – |
| 5 | Claude Opus 4.6 (unknown thinking) API access | Anthropic | 79,8 % | – |
| 6 | GPT-5.3 Codex API access | OpenAI | 78,4 % | – |
| 7 | Claude Opus 4.6 (no thinking) API access | Anthropic | 69,9 % | – |
| 8 | Gemini 3 Pro Preview API access | Google DeepMind | 69,4 % | – |
| 9 | GPT-5.2 Codex API access | OpenAI | 66,5 % | – |
| 10 | GPT-5.2 (unknown thinking) API access | OpenAI | 64,9 % | – |
| 11 | GPT-5.2 (medium) API access | OpenAI | 64,9 % | – |
| 12 | Gemini 3 Flash Preview API access | Google DeepMind | 64,3 % | – |
| 13 | Claude Opus 4.5 (unknown thinking) API access | Anthropic | 63,1 % | – |
| 14 | Claude Opus 4.5 (no thinking) API access | Anthropic | 63,1 % | – |
| 15 | GPT-5.1 Codex Mini API access | OpenAI | 61,6 % | – |
| 16 | GPT-5.1-Codex-Max API access | OpenAI | 60,4 % | – |
| 17 | Claude Opus 4.5 (128k thinking) API access | Anthropic | 59,1 % | – |
| 18 | GPT-5.1 Codex API access | OpenAI | 57,8 % | – |
| 19 | grok-4-20 API access | xAI | 57,3 % | – |
| 20 | Claude Sonnet 4.6 (unknown thinking) API access | Anthropic | 53,4 % | – |
| 21 | Claude Sonnet 4.6 (no thinking) API access | Anthropic | 53,4 % | – |
| 22 | GLM-5 Open weights (unrestricted) | Z.ai (Zhipu AI) | 52,4 % | – |
| 23 | GPT-5 (unknown thinking) API access | OpenAI | 49,6 % | – |
| 24 | GPT-5 (medium) API access | OpenAI | 49,6 % | – |
| 25 | GPT-5.1 (medium) API access | OpenAI | 47,6 % | – |
| 26 | GPT-5.1 (unknown thinking) API access | OpenAI | 47,6 % | – |
| 27 | Claude Sonnet 4.5 (unknown thinking) API access | Anthropic | 46,5 % | – |
| 28 | MiniMax-M2.7 Open weights (non-commercial) | MiniMax | 45,1 % | – |
| 29 | GPT-5-codex API access | OpenAI | 44,3 % | – |
| 30 | Kimi K2.5 Open weights (unrestricted) | Moonshot | 43,2 % | – |
| 31 | Claude Sonnet 4.5 (no thinking) API access | Anthropic | 42,8 % | – |
| 32 | MiniMax-M2.5 Open weights (unrestricted) | MiniMax | 42,7 % | – |
| 33 | DeepSeek-V3.2 (Thinking; Novita) Open weights (unrestricted) | DeepSeek | 39,6 % | – |
| 34 | claude-opus-4-1-20250805 API access | Anthropic | 38 % | – |
| 35 | Claude Opus 4.1 (unknown thinking) API access | Anthropic | 38 % | – |
| 36 | MiniMax-M2.1 Open weights (restricted use) | MiniMax | 36,6 % | – |
| 37 | Kimi K2 Thinking Open weights (restricted use) | Moonshot | 35,7 % | – |
| 38 | Claude Haiku 4.5 (unknown thinking) API access | Anthropic | 35,5 % | – |
| 39 | GPT-5 mini (unknown thinking) API access | OpenAI | 34,8 % | – |
| 40 | GLM-4.7 Open weights (unrestricted) | Z.ai (Zhipu AI) | 33,4 % | – |
| 41 | gemini-2.5-pro API access | Google DeepMind | 32,6 % | – |
| 42 | GPT-5 mini (medium) API access | OpenAI | 31,9 % | – |
| 43 | MiniMax-M2 Open weights (unrestricted) | MiniMax | 30 % | – |
| 44 | claude-haiku-4-5-20251001 API access | Anthropic | 29,8 % | – |
| 45 | Kimi K2 Instruct Open weights (restricted use) | Moonshot | 27,8 % | – |
| 46 | grok-4-0709 API access | xAI | 27,2 % | – |
| 47 | Qwen3-Coder-480B-A35B-Instruct Open weights (unrestricted) | Alibaba | 27,2 % | – |
| 48 | grok-code-fast-1 | – | 25,8 % | – |
| 49 | Qwen3.6 35B-A3B Open weights (unrestricted) | Alibaba | 24,6 % | – |
| 50 | GLM-4.6 Open weights (unrestricted) | Z.ai (Zhipu AI),Tsinghua University | 24,5 % | – |
| 51 | GPT-5 nano (unknown thinking) API access | OpenAI | 21,8 % | – |
| 52 | gpt-oss-120b Open weights (unrestricted) | OpenAI | 18,7 % | – |
| 53 | gpt-oss-120b (unknown thinking) Open weights (unrestricted) | OpenAI | 18,7 % | – |
| 54 | gemini-2.5-flash-preview-09-2025 API access | Google DeepMind | 17,1 % | – |
| 55 | gemini-2.5-flash API access | Google DeepMind | 17,1 % | – |
| 56 | GPT-5 nano (medium) API access | OpenAI | 11,5 % | – |
| 57 | Qwen3.5 9B Open weights (restricted use) | Alibaba | 9,2 % | – |
| 58 | gpt-oss-20b Open weights (unrestricted) | OpenAI | 3,4 % | – |
| 59 | gpt-oss-20b Open weights (unrestricted) | OpenAI | 3,4 % | – |
