{"id":"tau-bench","name":"tau-bench and tau2-bench","maker":"Sierra","url":"https://taubench.com","repo":"sierra-research/tau2-bench","summary":"Benchmarks for tool-using agents that talk with a simulated user while following domain policies.","details":"Agents handle customer service tasks such as airline and retail requests, calling tools against a database while a language model plays the user. The score checks the final database state and whether the agent followed the policy. The second version adds further domains. The code is MIT licensed.","open_source":true,"hosted":false,"verified":"2026-11-15","category":"benchmarks","category_name":"Agent benchmarks","page":"https://modelmoot.com/tools/tau-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."}