Terminal-Bench Hard

An agentic benchmark evaluating AI capabilities in terminal environments through software engineering, system administration, and data processing tasks.

Models: 147 current, 432 in total
Median: 6.8%
Range: 0.0%–53.8%

Score distribution

Lowest: 0.0%
Median: 6.8%
Highest: 53.8%

Leaderboard

Include deprecated models

Data from Artificial Analysis