OpenAI launches GPT-4o Design: a model tuned for product and UX workflows

AI · 5 min read

OpenAI today announced GPT-4o Design, a specialized instance of its GPT-4o family tailored for product teams and UX workflows. The model is tuned on paired design artifacts and textual intent, which OpenAI says improves its ability to output component JSON, accessibility props, and layout suggestions from short prompts.

GPT-4o Design introduces layout-aware tokenization so the model can reason about spatial relationships and typical UI hierarchies. The release includes prebuilt prompt templates for extracting design tokens, generating microcopy for forms and modals, and translating wireframes into annotated component specs.

The company is shipping SDKs that convert model outputs into Figma JSON, React props, and SwiftUI snippets, plus a verification mode that flags potential accessibility issues. Early partners report faster handoffs between research, design, and engineering, though some teams caution about the need for rigorous prompt templates to avoid inconsistent component naming.