Benchmark
tau2-bench v1 retail pass^3
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models measured
- 7
- Scores recorded
- 8
- Sources
- 1
- IQ weight
- 0.08
Score distribution
- 56.1%
- Mean
- 57.9%
- Median
- 5.5%
- Std dev
- 51.1%
- 10th percentile
- 59.9%
- 90th percentile
- 47.4%
- Min
This benchmark has a weight of 0.08 for chat models, measured over 7 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 | Qwen3.5 397B A17Bsubmission-qwen3.5-397b-a17b-think_sierra_2026-03-02, usersim-gpt-5.2, effort-enabled· Sierra tau-bench Leaderboard data docs | 66.5% |
| 2 | GPT 5.2submission-gpt-5-2_sierra_2026-02-26, usersim-gpt-5.2, effort-high· Sierra tau-bench Leaderboard data docs | 59.9% |
| 3 | Claude Opus 4.5submission-claude-opus-4-5_sierra_2026-02-26, usersim-gpt-5.2, effort-high· Sierra tau-bench Leaderboard data docs | 58.8% |
| 4 | Gemini 3 Flashsubmission-gemini-3-flash_sierra_2026-03-02, usersim-gpt-5.2, effort-high· Sierra tau-bench Leaderboard data docs | 57.9% |
| 5 | Gemini 3 Pro Previewsubmission-gemini-3-pro_sierra_2026-03-02, usersim-gpt-5.2, effort-high· Sierra tau-bench Leaderboard data docs | 54.4% |
| 6 | GPT 5.2submission-gpt-5-2-none_sierra_2026-02-26, usersim-gpt-5.2, effort-none· Sierra tau-bench Leaderboard data docs | 53.1% |
| 7 | GLM 5submission-glm-5-think_sierra_2026-03-02, usersim-gpt-5.2, effort-enabled· Sierra tau-bench Leaderboard data docs | 51.1% |
| 8 | Claude Sonnet 4.5submission-claude-sonnet-4-5_sierra_2026-02-26, usersim-gpt-5.2, effort-enabled· Sierra tau-bench Leaderboard data docs | 47.4% |
Data credits: Sierra Research tau2-bench leaderboard submissions (github.com/sierra-research/tau2-bench), MIT (licence)