Skip to content

Benchmark

Tau-bench

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages.

Models measured
13
Scores recorded
13
Sources
2
IQ weight
0.00

Score distribution

75.4%
Mean
77.2%
Median
8.9%
Std dev
63.2%
10th percentile
87.4%
90th percentile
62.9%
Min
62.9%88.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.00 for chat models, measured over 13 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Step 3.5 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
88.2%
2GLM 4.7· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
87.4%
3MiMo V2 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
80.3%
4GLM 4.7 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
79.5%
5MiniMax M2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.
Values from 2 sources
77.2%
6Claude Opus 4.5· Steel.dev Agent Leaderboards data70.2%
7GPT 5.2· Steel.dev Agent Leaderboards data69.9%
8Qwen3.5 397B A17B· Steel.dev Agent Leaderboards data68.4%
9Gemini 3 Flash· Steel.dev Agent Leaderboards data67.8%
10Gemini 3 Pro Preview· Steel.dev Agent Leaderboards data65.8%
11GLM 5· Steel.dev Agent Leaderboards data63.2%
12OpenAI o3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim63%
13Claude Sonnet 4.5· Steel.dev Agent Leaderboards data62.9%

Data credits: Steel.dev Agent Leaderboards, MIT (licence) · Data by LLM Stats