Back to Benchmarks
🤖
Tau-Bench
Tau-squared Benchmark
Tool-augmented NLP tasks with pass@N evaluation
🤖 Agentic / Tool Use
Best: 80.3%
7 Models Evaluated
Champion
Claude Opus 4.6
Anthropic
91.9%
Best Score
Model Rankings
All models ranked by score on Tau-Bench. Scores are relative to the best score of 80.3%.
| # | Model | Provider | Score | vs Best |
|---|---|---|---|---|
| 1 | Claude Opus 4.6 Elo 1503 | Anthropic | 91.9% | Best |
| 2 | Claude Sonnet 4.6 Elo 1460 | Anthropic | 91.7% | +11.4% |
| 3 | Qwen 3.5 Elo 1450 | Alibaba | 86.7% | +6.4% |
| 4 | Gemini 3.1 Pro Elo 1492 | 85.3% | +5.0% | |
| 5 | MiMo-V2-Flash Elo 1393 | Xiaomi | 80.3% | +0.0% |
| 6 | Claude Opus 4.7 Elo 1500 | Anthropic | 70.2% | -10.1% |
| 7 | DeepSeek V4 Pro (Max) | DeepSeek | 67.9% | -12.4% |
Score Distribution
Claude Opus 4.6
91.9%
Claude Sonnet 4.6
91.7%
Qwen 3.5
86.7%
Gemini 3.1 Pro
85.3%
MiMo-V2-Flash
80.3%
Claude Opus 4.7
70.2%
DeepSeek V4 Pro (Max)
67.9%
Run This Benchmark
Learn how to evaluate your own models on Tau-Bench.
Tau-Bench is a standardized benchmark for evaluating AI model capabilities.
To run this benchmark on your own models, visit the official benchmark page or leaderboard for detailed instructions, code, and evaluation scripts.
Data compiled from official model cards, public leaderboards, and independent evaluations.