Skip to content

Benchmark

BrowserGym VisualWebArena

Vision benchmarks measure how well models understand images. Scores are percentages.

Models measured
3
Scores recorded
3
Sources
1
IQ weight
No weight published

Score distribution

21.5%
Mean
21%
Median
4%
Std dev
16.9%
10th percentile
26.7%
90th percentile
16.9%
Min
16.9%26.7%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1GPT 4oGenericAgent· BrowserGym Leaderboard26.7%
2Claude 3.5 SonnetGenericAgent· BrowserGym Leaderboard21%
3GPT-4o miniGenericAgent· BrowserGym Leaderboard16.9%

Data credits: BrowserGym Leaderboard (ServiceNow), MIT (licence)