Anthropic releases Claude-Design, a model tuned for human-centered UI prompts

AI · 5 min read

Anthropic announced Claude-Design, a model variant trained on UX research reports, persona documents, and annotated wireframe datasets to better understand intent-driven design prompts. The model outputs structured wireframes, user flows, and suggested usability tests that designers can run to validate concepts.

Claude-Design also includes built-in safety to avoid producing UI that violates accessibility heuristics or privacy best practices. Its output format is JSON-first, making it easy to plug into design systems or generate interactive prototypes.

Early access partners used the model to accelerate research-to-prototype cycles, noting improved alignment between designer intent and generated wireframes. Anthropic said it will iterate on the model with partner feedback and open a limited beta to product teams this quarter.