Anthropic launches Claude Creative with controllable style tokens for designers

AI · 4 min read

Anthropic released Claude Creative, a variant of its Claude family tuned for iterative creative work. The core feature is a style-token system that allows teams to define visual and verbal constraints — for example, "playful, 16px rounded, high-contrast" — which the model then applies across copy and UI suggestions.

Claude Creative supports multi-turn critique cycles where the model can act as a reviewer, pointing out accessibility issues, naming inconsistencies, and spacing anomalies. It can propose changes with inline rationale, enabling designers to accept, tweak, or reject recommendations in the same session.

Anthropic is pitching the model to in-house design teams and agencies that need reliable, brand-aligned generation. The company emphasizes safety and explainability: each suggestion is accompanied by a short chain-of-thought summary indicating why the model made certain trade-offs.