Benchmark
Tau2 Retail
communication
reasoning
text
τ²-bench retail domain evaluates conversational AI agents in customer service scenarios within a dual-control environment where both agent and user can interact with tools. Tests tool-agent-user interaction, rule adherence, and task consistency in retail customer support contexts.
语言EN
满分1
参评模型10
模型排名
| 名次 | 模型 | 机构 | 分数 | 来源 |
|---|---|---|---|---|
| 1 | Claude Haiku 4.5 | Anthropic | 83.2 | 来源 ↗ |
| 2 | GPT-5 | OpenAI | 81.1 | 来源 ↗ |
| 3 | o3 | OpenAI | 80.2 | 来源 ↗ |
| 4 | Qwen3-235B-A22B-Thinking-2507 | Alibaba Cloud / Qwen Team | 71.9 | 来源 ↗ |
| 5 | Qwen3-235B-A22B-Instruct-2507 | Alibaba Cloud / Qwen Team | 71.3 | 来源 ↗ |
| 6 | Kimi K2 Instruct | Moonshot AI | 70.6 | 来源 ↗ |
| 7 | Kimi K2-Instruct-0905 | Moonshot AI | 70.6 | 来源 ↗ |
| 8 | Qwen3-Next-80B-A3B-Thinking | Alibaba Cloud / Qwen Team | 67.8 | 来源 ↗ |
| 9 | GPT-4o | OpenAI | 63.4 | 来源 ↗ |
| 10 | Qwen3-Next-80B-A3B-Instruct | Alibaba Cloud / Qwen Team | 57.3 | 来源 ↗ |