Skip to content

Benchmark

Open Agent Leaderboard Tau Bench 2 retail

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

Models measured
5
Scores recorded
25
Sources
1
IQ weight
0.19

Score distribution

65.9%
Mean
73%
Median
22.9%
Std dev
19%
10th percentile
82%
90th percentile
1%
Min
1%85%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.19 for chat models, measured over 5 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Opus 4.5agent: openai_solo, task: tau2_retail, subset: retail, route: openai/aws· Open Agent Leaderboard results data docs85%
2Claude Opus 4.5agent: claude_code, task: tau2_retail, subset: retail, route: openai/aws· Open Agent Leaderboard results data docs83%
3DeepSeek V3.2agent: tool_calling_with_shortlisting, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs82%
4DeepSeek V3.2agent: tool_calling, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs82%
5Gemini 3 Pro Previewagent: tool_calling_with_shortlisting, task: tau2_retail, subset: retail, route: openai/gcp· Open Agent Leaderboard results data docs82%
6Gemini 3 Pro Previewagent: tool_calling, task: tau2_retail, subset: retail, route: openai/gcp· Open Agent Leaderboard results data docs82%
7Claude Opus 4.5agent: smolagents_code, task: tau2_retail, subset: retail, route: openai/aws· Open Agent Leaderboard results data docs78%
8Claude Opus 4.5agent: tool_calling_with_shortlisting, task: tau2_retail, subset: retail, route: openai/aws· Open Agent Leaderboard results data docs78%
9Claude Opus 4.5agent: tool_calling, task: tau2_retail, subset: retail, route: openai/aws· Open Agent Leaderboard results data docs78%
10DeepSeek V3.2agent: smolagents_code, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs77%
11Gemini 3 Pro Previewagent: smolagents_code, task: tau2_retail, subset: retail, route: openai/gcp· Open Agent Leaderboard results data docs75%
12Gemini 3 Pro Previewagent: openai_solo, task: tau2_retail, subset: retail, route: openai/gcp· Open Agent Leaderboard results data docs73%
13GPT 5.2agent: tool_calling_with_shortlisting, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs73%
14GPT 5.2agent: tool_calling, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs73%
15Kimi K2.5agent: smolagents_code, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs72.4%
16Gemini 3 Pro Previewagent: claude_code, task: tau2_retail, subset: retail, route: openai/gcp· Open Agent Leaderboard results data docs71%
17GPT 5.2agent: smolagents_code, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs68%
18DeepSeek V3.2agent: claude_code, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs65%
19Kimi K2.5agent: tool_calling_with_shortlisting, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs64.6%
20Kimi K2.5agent: tool_calling, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs64.6%
21GPT 5.2agent: claude_code, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs64%
22GPT 5.2agent: openai_solo, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs53%
23DeepSeek V3.2agent: openai_solo, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs19%
24Kimi K2.5agent: claude_code, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs3%
25Kimi K2.5agent: openai_solo, task: tau2_retail, subset: retail, route: openai/azure· Open Agent Leaderboard results data docs1%

Data credits: Open Agent Leaderboard results (Exgentic), CDLA-Permissive-2.0; agreement text at https://cdla.dev/permissive-2-0/. Fractions shown as percentages. (licence)