Skip to content

Benchmark

BrowserGym WorkArena-L1

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

Models measured
12
Scores recorded
12
Sources
1
IQ weight
1.76

Score distribution

48.4%
Mean
48.2%
Median
14.3%
Std dev
27.9%
10th percentile
63.3%
90th percentile
27%
Min
27%79.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.76 for chat models, measured over 12 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1GPT 5GenericAgent· BrowserGym Leaderboard79.1%
2Claude Sonnet 4GenericAgent· BrowserGym Leaderboard63.3%
3GPT-5 miniGenericAgent· BrowserGym Leaderboard60.6%
4Claude 3.5 SonnetGenericAgent· BrowserGym Leaderboard56.4%
5GPT OSS 120BGenericAgent· BrowserGym Leaderboard50.9%
6o3 MiniGenericAgent· BrowserGym Leaderboard48.2%
7GPT 4oGenericAgent· BrowserGym Leaderboard45.5%
8Llama 3.1 405BGenericAgent· BrowserGym Leaderboard43.3%
9GPT-5 nanoGenericAgent· BrowserGym Leaderboard40.6%
10GPT OSS 20bGenericAgent· BrowserGym Leaderboard38.5%
11Llama 3.1 70BGenericAgent· BrowserGym Leaderboard27.9%
12GPT-4o miniGenericAgent· BrowserGym Leaderboard27%

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