GPT-5.6 lands in Figma Make

Figma has added OpenAI’s GPT-5.6 to Figma Make, the company’s AI-powered build tool. Based on Figma’s early testing, the new model improves both the quality and speed of first-pass output, covering everything from UI cleanliness and responsive layouts to higher-fidelity interactions.

The upgrade is intended to reduce the distance between an initial prompt and a working prototype, allowing teams to validate more directions in a single work session. Figma’s evaluation runs suggest GPT-5.6 handles complex design requests well; one test involved generating a stock tracking app with a dark, gothic aesthetic from scratch. The output included an interactive dashboard with sample price values and performance figures, plus additional details such as keyboard shortcuts and search — all from a single prompt.

GPT-5.6 also demonstrates improved resilience when builds encounter issues. Rather than stalling, the model investigates errors and attempts to fix them on its own. In one instance, it identified the cause of a blank build and corrected it without further prompting, reducing the back-and-forth typically needed to reach a working result.

Fidelity to existing designs

The model also performs strongly when working from existing material, such as a design spec or a Figma Design file. In one test, a prompt based on a nature sound player app with a multi-track timeline, playback controls, and a sound library produced a first pass that preserved the original layout, visual hierarchy, spacing, proportions, and styling. Interactive elements functioned as intended, with play, pause, and skip controls working and audio files playing correctly.

The result is that generating an interactive prototype from a static design becomes a more dependable step in a design-to-code workflow.

Quality on the first pass

Figma also tested GPT-5.6 with a minimal prompt for a content-heavy e-commerce site featuring a bookshelf product catalog. The first pass came back clean and professional, including product descriptions, measurements, and care guides. Interactions worked immediately — dropdowns contained populated descriptions, the photo library was interactive, and the menu was clickable. The build also adapted across screen sizes without additional prompting.

GPT-5.6 is available now in the model selector within Figma Make. More details on working with the model can be found in Figma’s help center.