BrowserGym WorkArena L1: leaderboard
BrowserGym evaluation on WorkArena Level 1: enterprise web agent tasks in a realistic ServiceNow environment. Tests basic workflows.
Metric: Success Rate (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 79.1 | #98 |
| 2 | Claude Sonnet 4 | 63.3 | #199 |
| 3 | GPT-5 Mini | 60.6 | #186 |
| 4 | Claude 3.5 Sonnet | 56.4 | #350 |
| 5 | GPT-OSS-120B | 50.9 | #341 |
| 6 | O3 Mini | 48.2 | #272 |
| 7 | GPT-4o | 45.5 | #343 |
| 8 | Llama 3.1 405B | 43.3 | #466 |
| 9 | GPT-5 Nano | 40.6 | #425 |
| 10 | GPT-OSS-20B | 38.5 | #518 |
| 11 | Llama 3.1 70B | 27.9 | #592 |
| 12 | GPT-4o Mini | 27 | #599 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=browsergym-workarena-l1 · How It Works · Data refreshed daily, snapshot 2026-10-20.