Meta launches Llama 3.2 with vision and mobile models

On September 25, 2024, Meta launched Llama 3.2, with for the first time multimodal models (11B and 90B) that also understand images, plus lightweight text models (1B and 3B) for phones and edge devices.

AI on your device

The small models run locally, without cloud, which benefits privacy and speed. The large models added image understanding to the open Llama ecosystem.

Open and broad

With Llama 3.2, Meta broadened its open offering from data center to smartphone.