Benchmark
TAU-bench Airline
reasoning
communication
text
Part of τ-bench (TAU-bench), a benchmark for Tool-Agent-User interaction in real-world domains. The airline domain evaluates language agents' ability to interact with users through dynamic conversations while following domain-specific rules and using API tools. Agents must handle airline-related tasks and policies reliably.
语言EN
满分1
参评模型20
模型排名
| 名次 | 模型 | 机构 | 分数 | 来源 |
|---|---|---|---|---|
| 1 | Claude Sonnet 4.5 | Anthropic | 70.0 | 来源 ↗ |
| 2 | GLM-4.5-Air | Zhipu AI | 60.8 | 来源 ↗ |
| 3 | GLM-4.5 | Zhipu AI | 60.4 | 来源 ↗ |
| 4 | Claude Sonnet 4 | Anthropic | 60.0 | 来源 ↗ |
| 5 | Claude Opus 4 | Anthropic | 59.6 | 来源 ↗ |
| 6 | Claude 3.7 Sonnet | Anthropic | 58.4 | 来源 ↗ |
| 7 | Claude Opus 4.1 | Anthropic | 56.0 | 来源 ↗ |
| 8 | o1 | OpenAI | 50.0 | 来源 ↗ |
| 9 | GPT-4.5 | OpenAI | 50.0 | 来源 ↗ |
| 10 | GPT-4.1 | OpenAI | 49.4 | 来源 ↗ |
| 11 | o4-mini | OpenAI | 49.2 | 来源 ↗ |
| 12 | Qwen3-Next-80B-A3B-Thinking | Alibaba Cloud / Qwen Team | 49.0 | 来源 ↗ |
| 13 | Qwen3-235B-A22B-Thinking-2507 | Alibaba Cloud / Qwen Team | 46.0 | 来源 ↗ |
| 14 | Claude 3.5 Sonnet | Anthropic | 46.0 | 来源 ↗ |
| 15 | Qwen3-Next-80B-A3B-Instruct | Alibaba Cloud / Qwen Team | 44.0 | 来源 ↗ |
| 16 | GPT-4o | OpenAI | 42.8 | 来源 ↗ |
| 17 | GPT-4.1 mini | OpenAI | 36.0 | 来源 ↗ |
| 18 | o3-mini | OpenAI | 32.4 | 来源 ↗ |
| 19 | Claude 3.5 Haiku | Anthropic | 22.8 | 来源 ↗ |
| 20 | GPT-4.1 nano | OpenAI | 14.0 | 来源 ↗ |