New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
TAU-bench Retail benchmark
AI model leaderboard for the TAU-bench Retail benchmark. Compare how large language models score on TAU-bench Retail, see the full ranking, and understand what this AI benchmark measures. Claude Opus 4 currently leads with 81.4%. Retail customer-service track of τ-bench — multi-turn agentic tool-use.
Leaderboard
| # | Model | Organization | Score | Variant | Source |
|---|---|---|---|---|---|
| #1 | Claude Opus 4 | Anthropic | 81.4% | reported in GLM-4.5 paper | official ↗ |
| #2 | Claude Sonnet 4 | Anthropic | 80.5% | reported in GLM-4.5 paper | official ↗ |
| #3 | GLM 4.5 | Zhipu AI | 79.7% | official ↗ | |
| #4 | GLM-4.5-Air | Zhipu AI | 77.9% | official ↗ | |
| #5 | gemini-2.5-pro | Google DeepMind | 77% | reported in GLM-4.5 paper | official ↗ |
| #6 | Grok 4 | xAI | 76.5% | reported in GLM-4.5 paper | official ↗ |
| #7 | GPT-4.1 | OpenAI | 75.1% | reported in GLM-4.5 paper | official ↗ |
| #8 | Kimi K2 | Moonshot | 73.9% | reported in GLM-4.5 paper | official ↗ |
| #9 | o3 | OpenAI | 70.4% | reported in GLM-4.5 paper | official ↗ |
| #10 | gpt-oss-120b | OpenAI | 67.8% | high reasoning | official ↗ |
| #11 | MiniMax-M1-40k | MiniMax | 67.8% | official ↗ | |
| #12 | gpt-oss-120b | OpenAI | 67.8% | high | official ↗ |
| #13 | o4-mini | OpenAI | 65.6% | reported in GLM-4.5 paper | official ↗ |
| #14 | MiniMax-M1-80k | MiniMax | 63.5% | official ↗ | |
| #15 | EXAONE 4.0 (32B) | LG AI Research | 62.8% | reasoning | official ↗ |
| #16 | gpt-oss-20b | OpenAI | 54.8% | high reasoning | official ↗ |
| #17 | gpt-oss-20b | OpenAI | 54.8% | high | official ↗ |
| #18 | EXAONE 4.0 (1.2B) | LG AI Research | 28.1% | reasoning | official ↗ |
Frequently asked questions about TAU-bench Retail
What is the TAU-bench Retail benchmark?
Retail customer-service track of τ-bench — multi-turn agentic tool-use.
How is the TAU-bench Retail benchmark scored?
TAU-bench Retail is scored using the pass^1 (%) metric, where a higher score is better. The maximum achievable score is 100.000. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.
Which AI model scores highest on TAU-bench Retail?
As of the latest reported scores on GenAIList, Claude Opus 4 achieves the highest result on TAU-bench Retail with a score of 81.4%.
Is a higher TAU-bench Retail score better?
Yes. On TAU-bench Retail a higher score indicates better performance, so models near the top of the leaderboard are the strongest.