OpenAI launches Mosaic-XL: a multimodal design assistant API for teams

AI · 6 min read

OpenAI's Mosaic-XL is presented as a team-focused multimodal model that understands layout semantics, visual style constraints, and natural language instructions. Unlike single-output image generators, Mosaic-XL returns layered assets, suggested layout variants, and a provenance trail that tracks how prompts and edits produced each iteration.

The API includes presets for mobile, web, and responsive breakpoints, and supports brand token inputs (colors, type scales, spacing) so generated designs adhere to style systems. OpenAI also released a companion SDK with components for Figma and Framer that translate Mosaic's layered outputs into editable design elements.

For enterprises, Mosaic-XL offers workspace-level controls: access policies, protected tokens, and change auditing. OpenAI positions the product as a co-designer that speeds ideation while keeping human designers in the loop, not as a replacement for final visual decisions.