← τ-bench leaderboards

Airline · Standard · pass^2

τ-bench · 8 models

RankModelScore
1GPT-5.278.33
2Claude Opus 4.577.67
3Gemini 3 Flash76.33
4GLM-576
5Qwen3.5-397B-A17B75.67
6Gemini 3 Pro74.67
7Claude Sonnet 4.562
8GPT-5.235.6667