VIBE (MiniMax): leaderboard

Each model, run by MiniMax, builds an application from a written requirement, and an automatic verifier that MiniMax says uses Claude Code as its scaffold deploys it and checks that it runs, that its interactions work and how it looks; the score averages three runs over the Web, Simulation, Android, iOS and Backend subsets. MiniMax does not say how many tasks each model ran. Unlike Vals' Vibe Code Bench it covers native mobile and scientific-simulation apps and grades with a verifying agent.

Metric: Score (0-100). Source: github.com. Saturation forecast: Estimated already saturated. 5 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 4.590.7#161
2MiniMax-M2.188.6#573
3Claude Sonnet 4.585.2#258
4Gemini 3 Pro82.4#145
5MiniMax-M267.5#687

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=vibe-minimax · How It Works · Data refreshed daily, snapshot 2026-11-17.