WebArena — leaderboard
Self-hosted environment with 812 web tasks across e-commerce, forums, code repos, and CMS. Tests autonomous web navigation and interaction.
Metric: Success Rate (%). Source: docs.google.com. Status: years away from saturation. 47 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Median Human | 78.24 |
| 2 | DeepSeek V3.2 | 74.3 |
| 3 | GPT-4 | 20.2 |
| 4 | GPT-4 (0613) | 14.9 |
| 5 | GPT-4o (2024-05-13) | 13.1 |
| 6 | Gemini 1.0 Pro | 7.12 |
Interactive version: theaggregate.ai/benchmark?slug=webarena · How the rankings work · Data refreshed daily, snapshot 2026-07-22.