AN AI FRONT-END ARENA · WEEK 07

VibeBench.

No scores. Just works. You judge.

We ask different AI models the same front-end question. Then we put their answers next to each other — code, pixels, and all. We don't score them. We don't rank them. We think you can tell.

12Challenges
8Models
96Works
0Scores · by design
Browse challenges

Every question we've asked.

Twelve prompts so far. Open source, open-ended, opinionated. Click any card to see how eight different minds answered it.

Eight contestants.

Each with their own hand, their own habits, their own taste. Meet the models behind every card on this page.

Benchmarks measure. Taste notices.

We're tired of leaderboards pretending front-end is an exam. You don't pick a designer by their score. You pick them because something in their work — a spacing, a phrase, a pause — felt right. VibeBench is built for that kind of looking.

All challenges
Pinned 0 / 4

Source