WebDev Arena: leaderboard
Code Arena WebDev ranking from LMArena. Models generate web apps from prompts and are ranked by human preference votes; this project ingests the corresponding webdev_arena_external.csv feed from Epoch AI benchmark data.
Metric: Arena Score. Source: arena.ai. Status: years away from saturation. 119 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5.1 (Max) | 1765.37 |
| 2 | Claude Opus 5 (Max) | 1690.64 |
| 3 | Kimi K3 (Max) | 1673.7 |
| 4 | Qwen 3.8 Max | 1668.94 |
| 5 | Claude Opus 5 (High) | 1662.59 |
| 6 | Grok 4.6 (High) | 1628.84 |
| 7 | GPT-5.6 Sol (xHigh) | 1618.71 |
| 8 | GLM-5.3 Flash | 1604.43 |
| 9 | GLM-5.3 (Max) | 1598.6 |
| 10 | Qwen 3.8 27B | 1594.78 |
| 11 | GLM-5.2 (Max) | 1593.25 |
| 12 | Gemini 3.7 Flash (High) | 1587.04 |
| 13 | DeepSeek V4 Flash (0731) (High) | 1579.1 |
| 14 | Gemini 3.8 Flash (High) | 1567.33 |
| 15 | Claude Opus 4.8 (High) | 1563.49 |
Interactive version: theaggregate.ai/benchmark?slug=webdev-arena · How It Works · Data refreshed daily, snapshot 2026-09-05.