Meta launches Llama 3 Vision with on-device designer-friendly tooling

AI · 5 min read

Meta today announced Llama 3 Vision, a multimodal successor engineered for efficient on-device image and layout understanding. The model emphasizes low-latency inference on modern mobile and ARM-based laptops while preserving multimodal accuracy for compositional tasks.

The release includes a new Designer SDK exposing layout-aware embeddings, auto-suggested color harmonies, and component extraction meant to accelerate ideation. Meta says the SDK supports common design file formats and can export repeatable tokens for tools like Figma and Sketch.

Early tests by design teams show Llama 3 Vision generates high-quality component suggestions and can convert screenshots into editable vector assets with 60–80% fidelity. Meta will provide an open model checkpoint under a permissive license for research and commercial integration, with enterprise support available through its cloud platform.