TauC-Bench - Retail (Gemini 3 Flash User): leaderboard
Metric: pass^1 (times 100, so 0-100): mean per-task success, a task succeeding when the database state after the conversation equals the gold final state, on the 114 generated Retail tasks of TauC-Bench, run in the tau2-bench harness with default settings and a Gemini 3 Flash user simulator; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 | 59 |
| 2 | Claude Sonnet 4.6 | 54 |
| 3 | DeepSeek V3.1 | 47 |
| 4 | Gemini 2.5 Flash | 36 |
| 5 | Qwen 3 32B | 27 |
Interactive version: theaggregate.ai/benchmark?slug=tauc-bench-retail-gemini-3-flash-user · How It Works · Data refreshed daily, snapshot 2026-10-07.