OpenAI releases GPT-4.5 Vision-Edit with native Figma export

AI · 5 min read

OpenAI today launched GPT-4.5 Vision-Edit, a follow-up to its multimodal series focused on image editing workflows. The model is tuned to accept iterative natural language and sketch inputs, producing high-fidelity image edits and layered exports intended for design tools.

The standout feature is an official Figma export format that preserves layers, masks, and named components, allowing designers to continue work in Figma without rebuilding assets. The API also returns change diffs, so teams can programmatically apply or revert edits in design systems.

Early beta testers report faster iteration cycles on product visuals and marketing assets. Pricing includes a metered editing tier and a higher-throughput plan for studios, while OpenAI says it will roll out plugins for Adobe Photoshop and Affinity Photo in coming months.