{"id":"terminal-bench","name":"Terminal-Bench","maker":"Laude Institute and Stanford University","url":"https://www.tbench.ai","repo":"harbor-framework/terminal-bench","summary":"A benchmark that tests whether agents can finish hard tasks in a command-line environment, with a public leaderboard.","details":"Each task runs in a container with a test script that checks the final state, covering work such as building software, system administration and data processing. The tasks and harness are open source under Apache 2.0 and the site lists results for agents and models.","open_source":true,"hosted":false,"also":["coding-agents"],"verified":"2026-11-15","category":"benchmarks","category_name":"Agent benchmarks","page":"https://modelmoot.com/tools/terminal-bench","reviewed":"2026-11-15","checked":{"at":"2026-11-15T16:10:15.342Z","address_answers":true},"note":"Descriptions are the site's own. Facts under \"checked\" were read automatically from the tool's address and public repository."}