← Benchmarks
Tau2 Telecom
τ²-Bench telecom domain evaluates conversational agents in a dual-control environment modeled as a Dec-POMDP, where both agent and user use tools in shared telecommunications troubleshooting scenarios that test coordination and communication capabilities.
id tau2-telecom · max 1 · 36 models reported
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 Anthropic | 0.99 |
| 2 | LongCat-Flash-Thinking-2601 Meituan · open | 0.99 |
| 3 | GPT-5.4 OpenAI | 0.99 |
| 4 | GPT-5.2 OpenAI | 0.99 |
| 5 | Claude Opus 4.5 Anthropic | 0.98 |
| 6 | GPT-5.5 OpenAI | 0.98 |
| 7 | Claude Sonnet 4.6 Anthropic | 0.98 |
| 8 | MiMo-V2-Pro Xiaomi | 0.97 |
| 9 | GPT-5 OpenAI | 0.97 |
| 10 | GPT-5.1 OpenAI | 0.96 |
| 11 | GPT-5.1 Instant OpenAI | 0.96 |
| 12 | GPT-5.1 Thinking OpenAI | 0.96 |
| 13 | GPT-5.4 mini OpenAI | 0.93 |
| 14 | Nova 2 Pro Amazon | 0.93 |
| 15 | GPT-5.4 nano OpenAI | 0.93 |