Figma introduces Generative Layout Engine and Native Model Hosting for teams

Design · 5 min read

Figma's Generative Layout Engine (GLE) can propose full-page layouts from a brief, respecting design tokens and component libraries. Designers can request variations, tweak constraints like grid systems and spacing scales, and lock parts of a frame while the model regenerates the rest.

Separately, Figma announced native model hosting for enterprise customers, which lets teams upload and serve fine-tuned models in a private, sandboxed environment. That removes the need to route proprietary brand assets through third-party APIs and enables consistent on-premise workflows for generation and assistant features.

The combination of GLE and in-app model hosting is meant to streamline concepting phases. Early testers appreciated the token-aware proposals that matched existing component libraries, though some flagged occasional odd spacing in nested components—something Figma says they'll refine with A/B datasets and designer feedback.