Frontier AI models still trail human designers by 27 points on UI creativity

The Emperor Pelican Has No Clothes

Kombai's Design Creativity Bench evaluates 16 frontier models on 80 open-ended UI tasks, measuring originality, creative range, and appropriateness. The top model, OpenAI's GPT-6 Astra, scores 65.5 against a human reference of 92.3, exposing a substantial gap in creative design decisions despite rapid progress in generating functional websites and product UIs. The full leaderboard, paper, datasets, and reproducible framework are publicly released.