HAL Online Mind2Web — leaderboard
Princeton HAL leaderboard for Online Mind2Web: real-world web browsing tasks on live websites requiring navigation, form-filling, and multi-step interaction.
Metric: Accuracy (%). Source: hal.cs.princeton.edu. Status: saturation imminent. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 (Medium) | 42.33 |
| 2 | Claude Sonnet 4 (20250514) | 40 |
| 3 | Claude 3.7 Sonnet (High) | 39.33 |
| 4 | O3 (Medium) | 39 |
| 5 | Claude 3.7 Sonnet (20250219) | 38.33 |
| 6 | GPT-4.1 (2025-04-14) | 36.33 |
| 7 | DeepSeek V3 (0324) | 32.33 |
| 8 | O4 Mini (High) | 32 |
| 9 | O4 Mini (Low) | 31.67 |
| 10 | Gemini 2.0 Flash (001) | 29 |
| 11 | DeepSeek R1 | 25.33 |
Interactive version: theaggregate.ai/benchmark?slug=hal-online-mind2web · How the rankings work · Data refreshed daily, snapshot 2026-07-22.