Skip to content

Benchmark

WebArena-Verified

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

Models and systems measured
10
Scores recorded
12
Sources
2
IQ weight
1.25

Score distribution

66.5%
Mean
67%
Median
2.9%
Std dev
64.8%
10th percentile
69%
90th percentile
59%
Min
59%71.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.25 for chat models, measured over 9 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1Claude Opus 4.8max · OpenClaw · DOM + screenshots · full 812 tasks · 150 tool calls · pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.71.2%
2Gemini 3.1 Pro Previewhigh · OpenClaw · DOM + screenshots · full 812 tasks · 150 tool calls · pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.69%
3Muse Spark 1.1Meta Model API · xhigh · OpenClaw · DOM + screenshots · full 812 tasks · 150 tool calls · pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.69%
4Claude Sonnet 4.6hard subset · 258 tasks · official prompt/grader · pass@1 · Average@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.67.7%
5GPT 5.4DOM + screenshot interaction· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.67.3%
6GPT 5.5xhigh · OpenClaw · DOM + screenshots · full 812 tasks · 150 tool calls · pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.67%
7Qwen3.8 MaxQwen3.8-Max release vision/video table · official WebArena grader · OSWorld scaffold· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.66.8%
8Claude Sonnet 4.6full set · official prompt/grader · pass@1 · Average@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.66.4%
9GPT 5.2DOM + screenshot interaction· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.65.4%
10Qwen 3.8 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim64.8%
11Qwen 3.8 27BQwen3.8-27B model card · official WebArena grader · OSWorld scaffold· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.64.8%

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
Muse Spark (version unspecified)Meta Model API · highest effort · version unspecified · OpenClaw · DOM + screenshots · full 812 tasks · 150 tool calls · pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.59%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats