OpenAI releases Muse 2: a multimodal creative model tailored for UX and product design

AI · 5 min read

OpenAI today announced Muse 2, an update to its creative-model lineup specifically trained on UI layouts, iconography, microcopy, and interaction patterns. The model supports text, image, and sketch inputs and produces layered outputs that map to design tool components rather than flat raster images.

Muse 2's standout features include a Figma plugin that exports components, tokens, and accessibility annotations directly into a file, and a Sketch-like JSON schema for teams that want deterministic re-creation. OpenAI says the model can preserve brand tokens and is tuned to prefer production-ready assets over generic variations.

OpenAI also introduced usage controls and a commercial license optimized for design teams, plus a model fine-tuning pipeline so agencies can inject their brand voice and component library. Early adopters report faster ideation cycles but warn that teams still need strict review processes for accessibility and interaction fidelity.