Tau2-Bench Telecom: leaderboard
Dual-control conversational AI benchmark simulating telecom support scenarios where agent and user coordinate actions to resolve service issues.
Metric: Success Rate (%). Source: artificialanalysis.ai. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 424 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-4.7 Flash (Reasoning) | 98.83 | #509 (GLM-4.7 Flash) |
| 2 | GLM-5-Turbo | 98.54 | #119 |
| 3 | Step 3.7 Flash | 98.54 | #236 |
| 4 | GLM-5 (Thinking) | 98.25 | #141 (GLM-5) |
| 5 | Qwen 3.6 Plus | 97.66 | #120 |
| 6 | GLM-5.1 (Thinking) | 97.66 | #97 (GLM-5.1) |
| 7 | Grok 4.3 (High) | 97.66 | #188 (Grok 4.3) |
| 8 | GLM-5 (Non-reasoning) | 97.37 | #141 (GLM-5) |
| 9 | GLM-5.1 (Non-reasoning) | 97.08 | #97 (GLM-5.1) |
| 10 | Grok 4.20 0309 (Reasoning) | 96.49 | #112 (Grok 4.20 0309) |
| 11 | DeepSeek V4 Pro (Reasoning, Max Effort) | 96.2 | #102 (DeepSeek V4 Pro) |
| 12 | Kimi K2.6 | 95.91 | #104 |
| 13 | Kimi K2.5 (Thinking) | 95.91 | #144 (Kimi K2.5) |
| 14 | Qwen 3.6 Max Preview | 95.91 | #85 |
| 15 | GLM-4.7 (Reasoning) | 95.91 | #192 (GLM-4.7) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=tau2-bench-telecom · How It Works · Data refreshed daily, snapshot 2026-10-20.