Vibe Code Bench v1.1: leaderboard
Vals AI benchmark for vibe-coding agents that build complete applications from product-style prompts and are scored on functional correctness and quality.
Metric: Accuracy (%). Source: www.vals.ai. Status: saturation imminent. 92 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5 (Max) | 90.35 |
| 2 | Claude Fable 5.1 (Max) | 90.26 |
| 3 | GPT-6 (Max) | 89.59 |
| 4 | Claude Opus 5 | 88.4 |
| 5 | Kimi K3 | 84.96 |
| 6 | Claude Opus 4.8 (Max) | 82.72 |
| 7 | DeepSeek V4 Pro (0813) (Max) | 82.3 |
| 8 | Claude Sonnet 5 (Max) | 81.33 |
| 9 | GPT-5.6 Sol (Max) | 80.5 |
| 10 | Muse Spark 1.2 (xHigh) | 79.1 |
| 11 | Gemini 3.8 Flash (High) | 78.65 |
| 12 | GLM-5.3 (Max) | 78.12 |
| 13 | Claude Opus 4.8 | 77.48 |
| 14 | GPT-5.6 Luna (Max) | 77.06 |
| 15 | Grok 4.6 (High) | 76.24 |
Interactive version: theaggregate.ai/benchmark?slug=vibe-code-bench-v1-1 · How It Works · Data refreshed daily, snapshot 2026-09-05.