τ³-bench average
Agents · 2026-03-18
τ³-bench evaluates multi-turn tool-using customer-service agents across airline, retail, telecom, and banking-knowledge domains. This row records the aggregate average when a source reports one score rather than domain-specific results.
Top models (higher is better)
| Model | Score |
|---|---|
| Macaron-V1-Venti | 69.3 |
| GLM-5.2 | 69.1 |
| Opus 4.8 | 67.7 |
| Gemini 3.1 Pro Preview | 67.1 |
| Qwen3.7-Max | 63.0 |
| MiniMax M3 | 61.2 |
| GPT-5.5 | 61.1 |