Vibe Code Bench v1.1
Code · 2026-03-02
Web-app generation benchmark testing whether models can build usable applications from natural-language prompts.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 90.4 |
| Claude Opus 5 | 88.4 |
| Kimi K3 | 85.0 |
| Opus 4.8 | 82.7 |
| Sonnet 5 | 81.3 |
| GPT-5.6 Sol | 80.5 |
| GPT-5.6 Luna | 77.1 |
| Muse Spark 1.1 | 72.2 |
| Opus 4.7 | 71.0 |
| GPT-5.5 | 69.8 |
| Grok 4.5 | 69.0 |
| GPT-5.6 Terra | 67.8 |
| GPT-5.4 | 67.4 |
| GLM-5.2 | 64.0 |
| GPT-5.3-Codex | 61.8 |
| Opus 4.6 | 57.6 |
| Gemini 3.6 Flash | 57.3 |
| Sonnet 4.6 | 55.8 |
| GPT-5.2 | 53.5 |
| DeepSeek-V4-Pro | 49.9 |
| Composer 2.5 | 49.6 |
| Gemini 3.5 Flash | 48.7 |
| GPT-5.4 Mini | 48.0 |
| Qwen3.7-Max | 47.7 |
| MiniMax M3 | 47.6 |
| Kimi K2.7 Code | 47.2 |
| Qwen3.7-Plus | 46.4 |
| MiMo-V2.5 | 42.2 |
| GPT-5.2-Codex | 37.9 |
| Kimi K2.6 | 37.9 |
| Gemini 3.5 Flash-Lite | 37.2 |
| MiMo-V2.5-Pro | 34.1 |
| Gemini 3.1 Pro Preview | 32.0 |
| GLM-5.1 | 31.5 |
| GPT-5.4 Nano | 26.1 |
| Qwen3.6 Plus (2026-04-02) | 25.6 |
| GPT-5.1 | 24.6 |
| GLM-5 | 23.4 |
| Sonnet 4.5 | 22.6 |
| GPT-5.1-Codex-Max | 22.2 |