OpenAI launches Muse-2: a multimodal design co‑pilot for rapid UI ideation

AI · 6 min read

OpenAI's Muse-2 arrives as a purpose-built multimodal model aimed specifically at product designers and UX teams. The release includes native plugins for Figma and Sketch, allowing designers to prompt the model with sketches, screenshots, or plain-language briefs and receive annotated mockups, copy refinements, and accessibility annotations.

Muse-2 emphasizes compositional reasoning: it generates wireframes aligned to platform conventions (iOS, Android, Web) and can propose responsive variants automatically. OpenAI says the model is trained on a mixture of design tokens, annotated UI screenshots, and human workflows to reduce hallucinated layout behavior and improve pixel-level consistency.

OpenAI is rolling Muse-2 out first to enterprise and design partners, with an API tier for tools to call for non-interactive batch tasks. The company also published guidelines and design datasets for safer usage, aiming to limit misuse of copyrighted UI patterns while fostering responsible model-assisted design workflows.