benchmark
Terminal-Bench
A benchmark for measuring AI agents on practical tasks completed in terminal environments.
What it is
Terminal-Bench defines containerized terminal tasks and evaluation infrastructure for reproducible agent comparisons.
Benchmark evidence
Terminal-Bench maintainers
Evaluated target: Published agent submissions
- Primary measure
- Task success
Limitation: Scores are specific to task set, harness version, model access, and agent configuration.
Methodology Results