Meta Connect 2024: Llama 3.2 brings open multimodal 11B and 90B to the race
Meta announced Llama 3.2, introducing multimodal capabilities to its 11B and 90B models for image analysis tasks. Meta also launched 1B and 3B models for edge devices. The 90B and 11B models will not be available in Europe, following EU regulators' concerns about Meta's data collection practices.
The release of Llama 3.2, which brings multimodality to some of the company's models, is the basis for most of the Meta AI announcements at Meta Connect 2024. Llama 3.2 11B and 90B are vision-enabled, which means they can process images to perform various tasks, including analyzing charts and graphs, captioning images, and identifying objects within images.
As suggested when Meta publicized that it would pause training its models on the public data of adult Facebook and Instagram users, neither of these models is available in Europe. However, in parallel with these two models, Meta is launching the very lightweight Llama 3.2 1B and 3B, which will be available in Europe and are meant to run locally on edge devices. Although the company has yet to announce a successor for the 405B model, it has done a remarkable job with the Llama 3.2 to cover all its bases and continue to pressure its competitors by providing open access to its products
