Benchmark
tau2-bench v0.1 telecom pass^1
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models measured
- 9
- Scores recorded
- 10
- Sources
- 1
- IQ weight
- 0.50
Score distribution
- 72%
- Mean
- 84.2%
- Median
- 23.8%
- Std dev
- 48.9%
- 10th percentile
- 98%
- 90th percentile
- 34%
- Min
This benchmark has a weight of 0.50 for chat models, measured over 9 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Qwen3 Maxsubmission-qwen3-max_qwen_2026-01-23, usersim-gpt-4.1-2025-04-14· Sierra tau-bench Leaderboard data docs | 98.2% |
| 2 | Claude Sonnet 4.5submission-claude-sonnet-4-5_anthropic_2025-10-02· Sierra tau-bench Leaderboard data docs | 98% |
| 3 | DeepSeek V3.2submission-deepseek-v3.2_deepseek_2025-12-01, usersim-deepseek-v3.2· Sierra tau-bench Leaderboard data docs | 96.2% |
| 4 | GPT 5submission-gpt-5_sierra_2025-08-09, usersim-gpt-4.1-2025-04-14· Sierra tau-bench Leaderboard data docs | 95.8% |
| 5 | Qwen3 Maxsubmission-qwen3-max_qwen_2024_09_23· Sierra tau-bench Leaderboard data docs | 84.2% |
| 6 | Kimi K2submission-kimi-k2_moonshot-ai_2025-07-11· Sierra tau-bench Leaderboard data docs | 65.8% |
| 7 | o4 Minisubmission-o4-mini_openai_2024-06-20, usersim-gpt-4.1-2025-04-14· Sierra tau-bench Leaderboard data docs | 50.2% |
| 8 | Claude Sonnet 3.7submission-claude-3-7-sonnet_anthropic_2024-06-20, usersim-gpt-4.1-2025-04-14· Sierra tau-bench Leaderboard data docs | 49% |
| 9 | GPT-4.1 minisubmission-gpt-4-1-mini_openai_2024-06-20, usersim-gpt-4.1-2025-04-14· Sierra tau-bench Leaderboard data docs | 48.9% |
| 10 | GPT 4.1submission-gpt-4-1_openai_2024-06-20, usersim-gpt-4.1-2025-04-14· Sierra tau-bench Leaderboard data docs | 34% |
Data credits: Sierra Research tau2-bench leaderboard submissions (github.com/sierra-research/tau2-bench), MIT (licence)