Anonymous AI model Gestalt tops AI design blind test, beats human reference work

Contra Labs released the Frontier Design Benchmark to test AI design skills on real commercial design tasks. The test had 33 tasks, 35 professional designers, five rounds of blind reviews, and 4,311 votes.

A stealth anonymous model code-named Gestalt ranked first with an Elo score of 1,931 and a 90% win rate. It beat the human reference work Human golden, Anthropic's Claude Fable 5.2, and OpenAI's GPT-6.1 Sol.

In direct matchups with human reference work, designers chose Gestalt 20 out of 28 times. In design quality scores, Gestalt got 83, Claude Fable 5.2 and Claude Opus 5.5 got 84, and the human reference work got 74.

Gestalt's AI-trace score was close to human work. Contra Labs has not revealed its team, model architecture, or identity, so its origin is still unconfirmed.

SponsoredDomoAI
AI Tools Mentioned

Comments

Loading comments…