Terminal-Bench

A benchmark for evaluating agents on realistic tasks in terminal environments, with task-specific tests to check completion.

Source: Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces.

The 2026 paper describes Terminal-Bench 2.0. The harness-design study evaluates version 2.1.

Referenced in An Empirical Study of Harness Design for Coding Agents.