τ³-Bench: Results as Published

tau3-bench tests a customer-service agent that talks with a simulated user and acts through domain tools while following a written policy. The repository includes airline, retail, telecom, and banking-knowledge domains, and this release also adds full-duplex voice. For the usual text domains, the reward is the product of two checks: the final database state must match the state produced by replaying a reference solution, and the agent must say the required facts. A single published tau3-bench number does not by itself say which domains were included.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.