OpenAI unveils Orion: a multimodal model tuned for design workflows

AI · 6 min read

OpenAI today announced Orion, a new multimodal model specifically tuned for design teams and product workflows. Orion blends image, vector, and layout understanding with a text-first API that can output Figma-ready components and SVGs. The company emphasized latency improvements and deterministic exports, aimed squarely at designers who need repeatable assets rather than varied creative outputs.

Orion includes a 'layout advisor' that takes wireframes or hand-drawn sketches and returns structured component trees, accessibility annotations, and spacing tokens. OpenAI provided an official Figma plugin preview that syncs generated components into a team's design system, preserving constraints and auto-generating responsive variations. Developers can also call Orion to produce optimized image assets and exportable SVGs within their CI pipelines.

Early adopters report substantial time savings in converting concepts into high-fidelity mocks, but some designers caution about over-reliance on suggested layouts. OpenAI acknowledged that Orion's training emphasized maintenance of brand guidelines and offered enterprise controls to lock styles. The company plans a staged rollout with enterprise sandboxing and expanded plugin support across major design tools.