Meta Releases LlamaX: A Multimodal Model for 3D Scene Layouts and UX Flow Design

Tech · 6 min read

Meta today announced LlamaX, a multimodal model trained to reason about 3D spaces and UX flows—particularly for AR/VR and spatial computing interfaces. The model accepts scene scans, floor plans, and interaction briefs and can propose UI overlays, interaction hotspots, and responsive placements that consider occlusion and ergonomics.

LlamaX is designed for mixed-reality prototyping: designers can upload photogrammetry or depth maps and receive overlay suggestions that adapt to user height, reach envelopes, and typical interaction patterns. Meta included examples demonstrating wayfinding overlays, context-aware tool palettes, and gaze-driven menus.

The company released an SDK for Unity and Unreal, emphasizing workflow integration for game and XR studios. Meta also published a transparency report outlining training data sources and opt-out mechanisms for creators whose content may have been used during training.