Agentic & Tool Use benchmark

# τ²-bench Retail leaderboard

> τ²-bench Retail results for 7 AI models, led by Qwen3.5 397B-A17B at 84.4%. What the benchmark measures, who runs it, and a source for every score.
- Canonical page: https://noometry.com/benchmarks/tau2-retail
- Last updated: 2026-10-10
- Title: τ²-bench Retail Leaderboard (October 2026): Scores by Model

As of October 2026, Qwen3.5 397B-A17B has the highest published τ²-bench Retail score on Noometry at 84.4%, out of 7 models with results.

Last verified October 10, 2026

## About τ²-bench Retail

Retail customer-service conversations (returns, exchanges, order changes) handled with tools under a written policy.

- **Category:** [Agentic & Tool Use](https://noometry.com/best/agentic)
- **Introduced:** 2025
- **Format:** Tool use with a simulated user
- **Unit:** Percent (random guessing ≈ 0%)
- **Official site:** [taubench.com](https://taubench.com/)

## Top 7 models

Top models on τ²-bench Retail

1.  Qwen3.5 397B-A17B 84.4%
2.  GPT-5.2 81.6%
3.  Claude Opus 4.5 79.6%
4.  Gemini 3 Flash Preview 76.8%
5.  Gemini 3 Pro 75.9%
6.  GLM-5 73.7%
7.  Claude Sonnet 4.5 72.4%
8.  6570758085

Sponsored placements are available on pages like this one. [Advertise on Noometry](https://noometry.com/advertise)

## All results

τ²-bench Retail results by model
| # | Model | Provider | Score | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | [Qwen3.5 397B-A17B](https://noometry.com/models/qwen3-5-397b-a17b) |  [![](/logos/alibaba.svg) Alibaba (Qwen)](https://noometry.com/providers/alibaba) | 84.4% | enabled | [τ²-bench](https://taubench.com/) | 2026-03-02 |
| 2 | [GPT-5.2](https://noometry.com/models/gpt-5-2) | [OpenAI](https://noometry.com/providers/openai) | 81.6% | high | [τ²-bench](https://taubench.com/) | 2026-02-26 |
| 3 | [Claude Opus 4.5](https://noometry.com/models/claude-opus-4-5) | [Anthropic](https://noometry.com/providers/anthropic) | 79.6% | high | [τ²-bench](https://taubench.com/) | 2026-02-26 |
| 4 | [Gemini 3 Flash Preview](https://noometry.com/models/gemini-3-flash-preview) |  [![](/logos/google.svg) Google](https://noometry.com/providers/google) | 76.8% | high | [τ²-bench](https://taubench.com/) | 2026-03-02 |
| 5 | [Gemini 3 Pro](https://noometry.com/models/gemini-3-pro) |  [![](/logos/google.svg) Google](https://noometry.com/providers/google) | 75.9% | high | [τ²-bench](https://taubench.com/) | 2026-03-02 |
| 6 | [GLM-5](https://noometry.com/models/glm-5) | [Z.ai (Zhipu)](https://noometry.com/providers/zai) | 73.7% | enabled | [τ²-bench](https://taubench.com/) | 2026-03-02 |
| 7 | [Claude Sonnet 4.5](https://noometry.com/models/claude-sonnet-4-5) | [Anthropic](https://noometry.com/providers/anthropic) | 72.4% | enabled | [τ²-bench](https://taubench.com/) | 2026-02-26 |

## Compare the leaders

-   [Qwen3.5 397B-A17B vs GPT-5.2](https://noometry.com/compare/gpt-5-2-vs-qwen3-5-397b-a17b)
-   [Qwen3.5 397B-A17B vs Claude Opus 4.5](https://noometry.com/compare/claude-opus-4-5-vs-qwen3-5-397b-a17b)
-   [Qwen3.5 397B-A17B vs Gemini 3 Flash Preview](https://noometry.com/compare/gemini-3-flash-preview-vs-qwen3-5-397b-a17b)
-   [Qwen3.5 397B-A17B vs Gemini 3 Pro](https://noometry.com/compare/gemini-3-pro-vs-qwen3-5-397b-a17b)
-   [GPT-5.2 vs Claude Opus 4.5](https://noometry.com/compare/claude-opus-4-5-vs-gpt-5-2)
-   [GPT-5.2 vs Gemini 3 Flash Preview](https://noometry.com/compare/gemini-3-flash-preview-vs-gpt-5-2)

## Other agentic & tool use benchmarks

-   [Terminal-Bench](https://noometry.com/benchmarks/terminal-bench)
-   [APEX-Agents](https://noometry.com/benchmarks/apex-agents)
-   [Berkeley Function Calling Leaderboard](https://noometry.com/benchmarks/bfcl)
-   [OSWorld 2.0](https://noometry.com/benchmarks/osworld-2)
-   [GDPval](https://noometry.com/benchmarks/gdpval)
-   [Remote Labor Index](https://noometry.com/benchmarks/remote-labor-index)
-   [TheAgentCompany](https://noometry.com/benchmarks/the-agent-company)
-   [τ²-bench Airline](https://noometry.com/benchmarks/tau2-airline)
-   [τ²-bench Banking](https://noometry.com/benchmarks/tau2-banking)
-   [τ²-bench Telecom](https://noometry.com/benchmarks/tau2-telecom)
-   [Cybench](https://noometry.com/benchmarks/cybench)
-   [DeepResearch Bench](https://noometry.com/benchmarks/deepresearch-bench)

## Frequently asked questions

### What does τ²-bench Retail measure?

Retail customer-service conversations (returns, exchanges, order changes) handled with tools under a written policy.

### Which model has the highest τ²-bench Retail score?

As of October 2026, Qwen3.5 397B-A17B has the highest published τ²-bench Retail score on Noometry at 84.4%, out of 7 models with results.

### What is the best open-weight model on τ²-bench Retail?

Qwen3.5 397B-A17B has the highest τ²-bench Retail accuracy among open-weight models at 84.4%, ranking 1 of 7 overall.

### Cite this page

Noometry. (2026). τ²-bench Retail leaderboard. Retrieved October 10, 2026, from https://noometry.com/benchmarks/tau2-retail

Quote Noometry with a link back to this page. It is also available in [Markdown](https://noometry.com/md/benchmarks/tau2-retail.md).