Terminal-Bench 4.0 — Science
Agents · 2026-09-22
Resolution rate on the 14 science tasks in Terminal-Bench 4.0.
Top models (higher is better)
| Model | Score |
|---|---|
| Kimi K3.1 Pro | 73.8 |
| Gemini 4 Carbon | 59.5 |
| GPT-6 Astra | 57.1 |
| Muse Wonder Full | 57.1 |
| GPT-6.1 Sol | 54.8 |
| GPT-6.1 Terra | 54.8 |
| Gemini 4 Argon | 52.4 |
| Muse Wonder | 47.6 |
| GPT-5.6 Sol | 42.9 |
| GPT-6 Sol | 42.9 |
| Muse Spark 1.3 | 40.5 |
| Opus 4.8 | 38.1 |
| DeepSeek V4.1 Flash 1017 | 38.1 |
| GLM-5.3 | 38.1 |
| GLM-5.3 Flash | 33.3 |
| MiMo-V2.6-Pro | 33.3 |
| Step 5 Preview | 33.3 |
| Grok 4.7 | 28.6 |
| DeepSeek-V4-Flash-0731 | 26.2 |
| GPT-5.6 Terra | 26.2 |
| MiniMax M3.1 Flash | 26.2 |
| Qwen3.8 Max (0902) | 26.2 |
| Claude Haiku 5.5 | 21.4 |
| DeepSeek V4.1 Flash | 21.4 |
| Gemini 3.8 Flash | 21.4 |
| GPT-6 Luna | 21.4 |
| MiMo-V2.6-Flash | 21.4 |
| DeepSeek V4 Pro 0813 | 19.0 |
| Ember-1 | 16.7 |
| Grok 4.6 | 16.7 |
| Mistral Large 4 Preview | 16.7 |
| GPT-5.6 Luna | 14.3 |
| Grok 4.5 | 14.3 |
| Sonnet 5 | 11.9 |
| Gemini 3.5 Flash | 11.9 |
| Gemini 3.7 Flash | 11.9 |
| Muse Spark 1.2 | 11.9 |
| Kimi K3 | 9.5 |
| Muse Spark Open | 7.1 |
| Gemini 3.6 Flash | 7.1 |