Terminal-Bench Science
Harbor / Professional work · Completing scientific research tasks in a terminal
Across e/accUnder observation
Evidence budgetNot scored
Upstream data timeUnverified
Last successful syncCollection not started
What it measures and how
Distinguish systems such as Claude Code, Codex and mini-SWE. Different system scores cannot directly represent model capability.
How this evidence is used
Under observation: 12 models with mixed runtime systems; coverage of new frontier models remains incomplete.
Limitations and attribution
Under observation: 12 models with mixed runtime systems; coverage of new frontier models remains incomplete.
Data license: Leaderboard data usage terms remain to be confirmed.
Results are published by Harbor. Raw scores and the e/acc consensus index use different scales and cannot be added directly.