Vibe Code Bench v1.1 (Vals Index 22-task subset)
Code · 2026-07-01
Vibe Code Bench v1.1 evaluated on the 22 representative app-building tasks selected for the Vals Index across UI, data, and workflow patterns. It remains a protocol-specific subset of the full OpenHands-based benchmark.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 93.5 |
| Kimi K3 | 91.3 |
| Opus 4.8 | 89.9 |
| GPT-5.6 Sol | 89.3 |
| Sonnet 5 | 86.9 |
| Claude Opus 5 | 85.7 |
| GPT-5.6 Luna | 81.1 |
| Opus 4.7 | 77.7 |
| GPT-5.5 | 77.1 |
| GLM-5.2 | 72.9 |
| GPT-5.6 Terra | 72.8 |
| Grok 4.5 | 72.8 |
| Muse Spark 1.1 | 72.4 |
| Gemini 3.6 Flash | 57.9 |
| MiniMax M3 | 56.5 |
| Sonnet 4.6 | 56.2 |
| DeepSeek-V4-Pro | 55.9 |
| Gemini 3.5 Flash | 54.7 |
| Qwen3.7-Max | 52.9 |
| MiMo-V2.5 | 48.1 |
| Qwen3.7-Plus | 46.9 |
| GPT-5.4 Mini | 42.5 |
| Kimi K2.6 | 42.3 |
| MiMo-V2.5-Pro | 36.3 |
| Gemini 3.1 Pro Preview | 34.2 |
| GLM-5.1 | 32.8 |
| Gemini 3.5 Flash-Lite | 32.8 |
| GPT-5.4 Nano | 30.0 |
| Qwen3.6 Plus (2026-04-02) | 28.3 |
| Gemini 3 Flash Preview | 22.0 |
| Inkling-Small | 19.0 |
| Grok 4.3 | 15.5 |
| Inkling | 13.3 |
| MiniMax M2.7 | 11.7 |
| Haiku 4.5 | 9.9 |
| Nemotron 3 Ultra 550B A55B | 7.1 |
| Laguna XS.2 | 5.5 |
| Laguna M.1 | 5.3 |
| Grok 4.20 | 1.5 |
| Mistral Medium 3.5 | 1.2 |