Terminal-Bench Science — Earth Sciences (Vals)
Science · 2026-09-23
Eight earth-science tasks under Vals' Terminus 2 protocol with infrastructure/provider-error retries.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 62.5 |
| Claude Opus 5.5 | 50.0 |
| Claude Fable 5.1 | 37.5 |
| Opus 4.8 | 25.0 |
| Claude Opus 5 | 25.0 |
| Claude Fable 5 | 12.5 |
| Gemini 3.5 Flash | 12.5 |
| GPT-5.6 Sol | 12.5 |
| GPT-5.6 Terra | 12.5 |
| Grok 4.7 | 12.5 |
| MiMo-V2.6-Pro | 12.5 |
| Muse Spark 1.3 | 12.5 |
| Sonnet 5 | 0.0 |
| DeepSeek V4.1 Flash | 0.0 |
| DeepSeek-V4-Flash-0731 | 0.0 |
| DeepSeek V4 Pro 0813 | 0.0 |
| Gemini 3.6 Flash | 0.0 |
| Gemini 3.7 Flash | 0.0 |
| Gemini 3.8 Flash | 0.0 |
| GLM-5.3 Flash | 0.0 |
| GLM-5.3 | 0.0 |
| GPT-5.6 Luna | 0.0 |
| Grok 4.6 | 0.0 |
| Hy4 Preview | 0.0 |
| Kimi K3 | 0.0 |
| Mercury 2.5 | 0.0 |
| MiMo-V2.6-Flash | 0.0 |
| Qwen3.8 27B | 0.0 |