OpenFrame 2.0 launches with layout-aware multimodal encoder for designers

AI · 5 min read

OpenFrame 2.0, an open-source multimodal model optimized for UI and layout tasks, was released today with a new layout-aware encoder that better understands page structure and component hierarchy.

According to the maintainers, the model was trained on a curated corpus of design files, screenshots, and their DOM-like annotations to capture spatial relationships and responsive behavior, allowing it to suggest component reflows and accessibility improvements.

Design tooling vendors have already started shipping plugins that call OpenFrame 2.0 for tasks such as auto-generating responsive variants, extracting CSS-like style sheets from mockups, and producing alt text for complex interfaces. Early benchmarks show a 22% improvement in layout coherence metrics versus the previous release.