Which AI model scores highest on τ²-bench?

Gemini 3.1 Pro by Google holds the best τ²-bench (retail) result among tracked models, at 90.8% (released Feb 19 2026). Higher scores are better on this benchmark.

What are the top 5 models on τ²-bench?

1. Gemini 3.1 Pro (Google) — 90.8%; 2. Gemini 3.0 Pro (Google) — 85.3%; 3. GPT-5.2 (OpenAI) — 82%; 4. GPT-5 (OpenAI) — 81.1%; 5. GPT-5 mini (OpenAI) — 78.3%.

How many models have a published τ²-bench score?

7 tracked models have a published τ²-bench (retail) score. Scores are collected from published lab reports and benchmark sources, with recorded sources credited alongside the results. We do not re-run the evaluations.

Agentic tasks

τ²-benchretail

task success in the retail domain of the second-generation customer-service evaluation

retail airline telecom telecom, adjusted prompt

Rankings

Higher is better

τ²-bench — frequently asked questions

: task success in the retail domain of the second-generation customer-service evaluation
: Gemini 3.1 Pro by Google holds the best τ²-bench (retail) result among tracked models, at 90.8% (released Feb 19 2026). Higher scores are better on this benchmark.
: 1. Gemini 3.1 Pro (Google) — 90.8%; 2. Gemini 3.0 Pro (Google) — 85.3%; 3. GPT-5.2 (OpenAI) — 82%; 4. GPT-5 (OpenAI) — 81.1%; 5. GPT-5 mini (OpenAI) — 78.3%.
: 7 tracked models have a published τ²-bench (retail) score. Scores are collected from published lab reports and benchmark sources, with recorded sources credited alongside the results. We do not re-run the evaluations.