Terminal-Bench-Science 0.1
Agents
Terminal-Bench-Science 0.1.0 comprises 70 expert-reviewed research workflows across life, physical, mathematical, engineering and earth sciences. Artificial Analysis runs the complete release with mini-swe-agent and reports pass@1 averaged over three repeats per task; a task passes only if all of its tests pass.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 68.1 |
| Gemini 4 Carbon | 64.8 |
| Claude Opus 5.5 | 63.3 |
| GPT-6.1 Sol | 58.1 |
| GPT-6.1 Terra | 57.6 |
| Claude Sonnet 5.5 | 53.3 |
| Gemini 4 Argon | 52.9 |
| Claude Fable 5.1 | 52.6 |
| Kimi K3.1 Pro | 38.1 |
| Muse Wonder Full | 37.1 |
| GPT-6 Sol | 30.0 |
| Claude Opus 5 | 28.6 |
| Muse Wonder | 25.7 |
| GPT-5.6 Sol | 22.4 |
| Claude Haiku 5.5 | 20.0 |
| GPT-5.6 Terra | 13.8 |
| DeepSeek V4.1 Flash 1017 | 12.9 |
| Qwen3.8 Max (0902) | 11.9 |
| Muse Spark 1.3 | 11.0 |
| Gemini 3.8 Flash | 10.5 |
| GLM-5.3 | 9.5 |
| Grok 4.7 | 9.5 |
| DeepSeek V4.1 Flash | 9.0 |
| GPT-6 Luna | 8.6 |
| Kimi K3 | 8.1 |
| Grok 4.6 | 6.2 |
| DeepSeek V4 Pro 0813 | 5.7 |
| MiMo-V2.6-Pro | 5.7 |
| GLM-5.3 Flash | 4.8 |
| Mistral Large 4 Preview | 4.8 |
| Ling 3.1 Flash 1019 | 4.3 |
| MiMo-V2.6-Flash | 4.3 |
| GPT-5.6 Luna | 3.3 |
| MiniMax M3.1 Flash | 3.3 |
| Step 5 Preview | 2.4 |
| Muse Spark Open | 1.9 |
| Qwen3.8 27B | 1.0 |
| Beam 1.0 501B A23B | 0.5 |
| MiniMax M3 | 0.5 |
| Command A+ (May 2026, BF16) | 0.0 |