Skip to content
Evaluation sources

Terminal-Bench Science

Harbor / Professional work · Completes scientific tasks in a terminal

Official evaluation
Across NexGateHub AI News Observing
Evidence budgetNot scored
Upstream data timeUnverified
Last successful syncCollection has not started

What it measures and how

Distinguishes Claude Code, Codex, mini-SWE, and other systems. Different-system results cannot directly represent model ability.

How this evidence is used

Under review: 12 models use mixed systems and current frontier coverage is incomplete.

Evaluation limits & attribution

Under review: 12 models use mixed systems and current frontier coverage is incomplete.

Data license: Leaderboard-data usage terms need confirmation.

Results are published by Harbor. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.