reasoning benchmark
Tau-bench Leaderboard
Tau-bench is a benchmark for tool-agent-user interaction in retail and airline domains.
Updated Sep 5, 2026
Models6
Model coverage6
MetricScore
EvidenceB
Tau-bench Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelST | Score88.20% | Percentile100.00% | Participants6 | EvidenceC | Evaluated |
| Rank02 | ModelZA | Score87.40% | Percentile80.00% | Participants6 | EvidenceC | Evaluated |
| Rank03 | ModelXI | Score80.30% | Percentile60.00% | Participants6 | EvidenceC | Evaluated |
| Rank04 | ModelZA | Score79.50% | Percentile40.00% | Participants6 | EvidenceC | Evaluated |
| Rank05 | ModelMI | Score77.20% | Percentile20.00% | Participants6 | EvidenceC | Evaluated |
| Rank06 | ModelOP | Score63.00% | Percentile0.00% | Participants6 | EvidenceC | Evaluated |
Tau-bench Highlights
The leading models and scores on this benchmark.
Tau-bench Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for Tau-bench
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is Tau-bench?
What Tau-bench measures and how its scores work.
Tau-bench tests language agents in dynamic conversations with users, API tools, and domain-specific policy guidelines.
It measures agents' ability to follow domain-specific rules and evaluates the consistency and reliability of their behavior over multiple trials using a Score reported as a ratio.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- Tau-bench
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about Tau-bench.
Which model scores highest on Tau-bench?
Step-3.5-Flash is currently ranked first with 88.20%.
What are the top three models on Tau-bench?
The current leaders are Step-3.5-Flash (88.20%), GLM-4.7 (87.40%), and MiMo-V2-Flash (80.30%).
Which Tau-bench model has the lowest official input price?
Step-3.5-Flash has the lowest matched official input price at $0.10 input / $0.30 output per 1M tokens.
Which models are fastest among Tau-bench results?
The fastest matched records are MiniMax M2 (70.00 tok/s via MiniMax), GLM-4.7-Flash (50.00 tok/s via ZAI), and o3 (50.00 tok/s via OpenAI).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.