Terminal-Bench 2.1
Agents · 2026-05-06
Terminal-Bench 2.1 is a revision of Terminal-Bench 2.0 that fixes issues in 28 of its 89 command-line tasks and introduces continuous validation. Scores are reported as average task accuracy.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Mythos Preview | 92.1 |
| DeepSeek V4.1 Flash | 90.6 |
| MiMo-V2.6-Pro | 89.9 |
| GPT-6 Astra | 89.9 |
| GPT-5.6 Sol | 89.5 |
| Gemini 3.8 Flash | 89.4 |
| Claude Opus 5 | 89.1 |
| Qwen3.8 Max (0902) | 88.8 |
| Grok 4.6 | 88.4 |
| Kimi K3 | 88.3 |
| GLM-5.3 | 88.2 |
| GPT-5.6 Terra | 88.0 |
| Claude Fable 5 | 88.0 |
| DeepSeek V4 Pro 0813 | 87.9 |
| MiMo-V2.6-Flash | 87.6 |
| Macaron-V1-Venti | 87.6 |
| Claude Sonnet 5.5 | 87.3 |
| Qwen3.8-Flash-Next | 86.1 |
| Gemini 3.7 Flash | 85.8 |
| GPT-6.1 Sol | 85.4 |
| GPT-6 Sol | 85.4 |
| Muse Spark 1.3 | 85.4 |
| Claude Haiku 5.5 | 85.0 |
| Grok 4.7 | 85.0 |
| Step 5 Preview | 85.0 |
| GPT-5.6 Luna | 84.7 |
| GPT-5.5 | 84.3 |
| GLM-5.3 Flash | 84.3 |
| Claude Opus 5.5 | 83.5 |
| Opus 4.8 | 82.7 |
| GLM-5.2 | 82.7 |
| Sonnet 5 | 82.4 |
| Fugu Ultra v1.0 | 82.1 |
| Qwen3.8 2.4T A95B | 82.0 |
| Ember-1 | 82.0 |
| Claude Fable 5.1 | 81.6 |
| Qwen3.8-Max | 81.3 |
| GPT-5.6 Sol Pro | 81.2 |
| Nanbeige4.3-33B | 80.9 |
| Muse Spark 1.2 | 80.1 |