Terminal-Bench-Science 0.1

The same command-line setup as Terminal-Bench, pointed at scientific work: the AI has to drive research tooling and computational workflows through to a result, rather than administer a machine. Version 0.1 is the first release of the task set, and scores run lower than on the general board. Higher is better.

GPT-6 Astra by OpenAI holds the best Terminal-Bench-Science 0.1 result among tracked models, at 64.6% (released Sep 3 2026). Higher scores are better on this benchmark.

1. GPT-6 Astra (OpenAI) — 64.6%; 2. Claude Fable 5.1 (Anthropic) — 52.6%; 3. Claude Opus 5 (Anthropic) — 29%; 4. Claude Fable 5 (Anthropic) — 24.7%.

4 tracked models have a published Terminal-Bench-Science 0.1 score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.