AA Terminal-Bench Science 0.1: leaderboard
Artificial Analysis evaluates all 70 tasks from Terminal-Bench Science 0.1.0 with mini-swe-agent, a 1,000-step cap and upstream task resource/time limits. A task passes only if every isolated-verifier test passes. Pass@1 averages three repeats per task; scores are percentages. The five domains contain 19 life, 17 physical, 17 mathematical, 9 engineering and 8 earth science tasks.
Metric: Pass@1 (%). Source: artificialanalysis.ai. Saturation forecast: Around April 2027. 100 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Carbon (High) | 64.76 | #2 (Gemini 4 Carbon) |
| 2 | GPT-6 Astra (Max) | 63.33 | #5 (GPT-6 Astra) |
| 3 | GPT-6 Astra (High) | 59.52 | #5 (GPT-6 Astra) |
| 4 | GPT-6.1 Sol (Max) | 58.1 | #7 (GPT-6.1 Sol) |
| 5 | GPT-6.1 Terra (Max) | 57.62 | #13 (GPT-6.1 Terra) |
| 6 | GPT-6 Astra (xHigh) | 54.76 | #5 (GPT-6 Astra) |
| 7 | GPT-6 Astra (Medium) | 53.33 | #5 (GPT-6 Astra) |
| 8 | Gemini 4 Argon (High) | 52.86 | #9 (Gemini 4 Argon) |
| 9 | GPT-6.1 Sol (xHigh) | 50.48 | #7 (GPT-6.1 Sol) |
| 10 | GPT-6.1 Sol (High) | 46.19 | #7 (GPT-6.1 Sol) |
| 11 | GPT-6.1 Terra (xHigh) | 41.9 | #13 (GPT-6.1 Terra) |
| 12 | GPT-6 Astra (Low) | 37.14 | #5 (GPT-6 Astra) |
| 13 | GPT-6.1 Sol (Medium) | 37.14 | #7 (GPT-6.1 Sol) |
| 14 | Muse Wonder Full (Max) | 37.14 | #11 (Muse Wonder Full) |
| 15 | Kimi K3.1 Pro | 34.29 | #16 |
No result here: #3 Claude Opus 5.5, #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=aa-terminal-bench-science-0-1 · How It Works · Data refreshed daily, snapshot 2026-10-29.