App-Bench: leaderboard
Benchmark evaluating how AI coding agents and web-app builders generate real web applications from one natural-language prompt without human edits.
Metric: Overall Score (%). Source: appbench.ai. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Orchids | 76.8 |
| 2 | Claude Code | 67.5 |
| 3 | v0 | 64.9 |
| 4 | Bolt | 53.6 |
| 5 | Google AI Studio | 50.3 |
| 6 | Codex | 38.4 |
| 7 | Replit | 35.1 |
| 8 | Cursor | 27.8 |
| 9 | Lovable | 25.8 |
| 10 | Gemini CLI | 0 |
Interactive version: theaggregate.ai/benchmark?slug=app-bench · How It Works · Data refreshed daily, snapshot 2026-09-05.