Meta has published the official model card and usage instructions for the Llama 3.2 90B Vision Instruct multimodal model, released on September 25, 2024.
The documentation provides integration guides for libraries such as Transformers, inference providers including vLLM and SGLang, and local deployment via Docker Model Runner. The release is governed by the Llama 3.2 Community License Agreement, which outlines redistribution rights and commercial terms based on monthly active user thresholds. Access to the model requires agreement to share contact information in accordance with Meta's Privacy Policy.