Benchmarks / Coding
Vibe Code Bench 1-100
Can models extend a working web application across a long sequence of dependent requests? Vals AI's overall score for Vibe Code Bench 1-100.
Vibe Code Bench 1-100
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Wonder Full | max | 34.0% |
| 2 | Opus 5.5 | xhigh | 30.4% |
| 3 | Opus 5 | max | 28.5% |
| 4 | Fable 5.1 | max | 28.0% |
| 5 | GPT-6 Astra | max | 27.6% |
| 6 | Wonder | max | 26.6% |
| 7 | GPT-5.6 Luna | max | 22.6% |
| 8 | Gemini 4 | high | 21.2% |
| 9 | Muse Spark 1.3 | max | 20.5% |
| 10 | Haiku 5.5 | max | 20.4% |
| 11 | GPT-5.6 Sol | max | 20.0% |
| 12 | GLM 5.3 | max | 20.0% |
| 13 | Gemini 3.8 Flash | high | 18.8% |
| 14 | Kimi K3 | max | 18.2% |
| 15 | DeepSeek V4 Pro 0813 | max | 17.5% |
| 16 | DeepSeek V4.1 Flash | high | 16.4% |
| 17 | GLM 5.3 Flash | max | 16.0% |
| 18 | GPT-5.6 Terra | max | 14.8% |
| 19 | Grok 4.6 | high | 14.8% |
| 20 | Mistral Large 4 | high | 14.3% |
| 21 | Sonnet 5 | max | 13.8% |
| 22 | Qwen 3.8 Max | max | 12.8% |
| 23 | MiniMax M3 | max | 9.2% |
| 24 | Inkling | max | 7.3% |
| 25 | Gemini 3.1 Pro | high | 6.7% |