OpenRouter Tau2-Bench Airline: leaderboard
OpenRouter's own τ²-Bench Airline run: 50 multi-turn airline customer-service tasks where the model must use tools and follow policy, executed against the production endpoints it serves.
Metric: Accuracy (%). Source: openrouter.ai. Saturation forecast: Around February 2027. 142 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 | 80 | #145 |
| 2 | Claude Fable 5 | 79.5 | #20 |
| 3 | Ling 3.1 Flash | 79.3 | #75 |
| 4 | Claude Fable 5.1 | 79.2 | #10 |
| 5 | Claude Opus 5 | 78.8 | #16 |
| 6 | Claude Sonnet 5.5 | 78.8 | #14 |
| 7 | Gemini 4 Argon | 78.7 | #9 |
| 8 | Qwen 3.5 397B A17B | 78.4 | #154 |
| 9 | Gemini 3.7 Flash | 78.3 | #24 |
| 10 | Gemini 3.5 Flash Lite | 78.3 | #174 |
| 11 | MiMo-V2.6-Flash | 77.9 | #74 |
| 12 | Qwen 3.8 27B | 77.5 | #123 |
| 13 | DeepSeek V4 Pro (0813) | 77.5 | #58 |
| 14 | GLM-5.3 | 77.3 | #44 |
| 15 | Qwen 3.5 122B A10B | 77.2 | #184 |
Interactive version: theaggregate.ai/benchmark?slug=openrouter-tau2-bench-airline · How It Works · Data refreshed daily, snapshot 2026-11-05.