Design Creativity Bench finds frontier AI models trail human designers on UI creativity

Kombai introduces Design Creativity Bench, which evaluates 16 frontier models on 80 open-ended UI design tasks across two product domains.

It scores originality, creative range, and appropriateness, and reports that the best model, GPT-6 Astra, reaches 65.5 against a human reference of 92.3.

The leaderboard, methodology, and evaluation framework are published on the project's website, arXiv, Hugging Face, and GitHub.