tau3-bench

Third-generation tau-bench customer-service agent simulations with policy following.

The best 2 · 2 models tested

Top 2 · best setting per model · 2 models, 2 results (0 independent)

Measured by independent testersIndependent Reported by the makerVendor-reported

tau3-bench leaderboard
				Thinking levelSetting
1	Qwen3.6 Plus	Qwen (Alibaba)	70.7%	Defaultthinking	Maker's own figureVendor-reportedQwen (Alibaba) ↗	—	2 Apr 2026
2	GLM-5.1	Z.ai (Zhipu)	70.6%	Defaultthinking	Maker's own figureVendor-reportedZ.ai ↗	—	7 Apr 2026