Web-Bench — leaderboard
Web development benchmark testing code generation for realistic web applications with Pass@1 evaluation across multiple frameworks.
Metric: Pass@1 (%). Source: huggingface.co. Status: saturation imminent. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4 (20250514) | 28 |
| 2 | Claude Opus 4 (20250514) (Thinking) | 25.6 |
| 3 | Gemini 2.5 Pro (Preview 06-05) | 25.3 |
| 4 | Claude 3.7 Sonnet (20250219) (Thinking) | 25.1 |
| 5 | Claude Sonnet 4 (20250514) | 25.1 |
| 6 | Claude Sonnet 4 (20250514) (Thinking) | 24.3 |
| 7 | Claude 3.5 Sonnet (20241022) | 23 |
| 8 | Claude 3.7 Sonnet (20250219) | 22.5 |
| 9 | Claude 3.5 Sonnet (20240620) | 22 |
| 10 | GPT-4.1 | 21.1 |
| 11 | GPT-4.1 Mini | 20.8 |
| 12 | Gemini 2.5 Pro (Preview 05-06) | 20.8 |
| 13 | GPT-4o | 17.2 |
| 14 | DeepSeek R1 0528 | 17.1 |
| 15 | DeepSeek Coder V2 | 16.7 |
Interactive version: theaggregate.ai/benchmark?slug=web-bench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.