Design Creativity Bench launches AI design creativity evaluation
What happened
On October 15, Hacker News featured Design Creativity Bench, a new benchmark for AI UI design creativity. It sets 80 open-ended UI tasks and evaluates 16 frontier models on originality, creative range and fit. OpenAI GPT-6 Astra scored best at 65.5; the human designer reference score is 92.3, leaving a clear gap between the best model and human reference. The leaderboard, methodology and reproducible evaluation framework are all public.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageWhy do LLM pelicans look like this?
新基准 Design Creativity Bench 用 80 个开放式 UI 任务、16 个前沿模型评测 AI 的设计创造力，从原创性、创意跨度和适配性三个维度打分。表现最好的 OpenAI GPT-6 Astra 仅得 65.5 分，人类设计师参考分为 92.3，差距明显。该基准的排行榜、方法论与可复现评测框架已公开。
Heat over time
Not enough continuous observations to draw a trend yet.