OpenAI releases Muse: on-device multimodal model tuned for designers

AI · 5 min read

OpenAI today announced Muse, a lightweight multimodal model optimized to run on local hardware and targeted at product and visual designers. Muse supports image generation, layout suggestions, and natural-language-to-component conversion, with an emphasis on tight iteration loops and offline privacy.

The company is shipping Muse with SDKs for Figma, Sketch, and desktop design apps, and claims the model can produce usable UI mocks and responsive layout variations in a few seconds on modern CPUs or discrete GPUs. OpenAI says the architecture trades some raw creative capacity for determinism, speed, and a much smaller memory footprint than cloud models.

Pricing combines a one-time licensing fee for the local runtime with optional cloud sync for asset versioning. OpenAI positions Muse as complementary to its larger cloud models, recommending Muse for day-to-day prototyping and reserving larger models for high-fidelity generation and heavy compute tasks.