Benchmark

Tau-bench

τ-bench: A benchmark for tool-agent-user interaction in real-world domains. Tests language agents' ability to interact with users and follow…

Modality
text
Categories
reasoning, general, agents, tool_calling
Openness
unknown
Source
llm_stats
Reported scores
6

Reported scores

llm_stats

ModelOrganizationReported valueReportedEvidence
GLM-4.7Z.ai0.8742025-12-22evidence
GLM-4.7-FlashZ.ai0.7952026-01-19evidence
MiMo-V2-FlashXiaomi0.8032025-12-16evidence
MiniMax M2MiniMax0.7722025-10-27evidence
o3OpenAI0.632025-04-16evidence
Step-3.5-FlashStepFun0.8822026-02-02evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.