Meta releases the open-source Lama 3.2 model, including image processing

After an overwhelming response to Llama 3.1 herd of models within two months since launch, including the 405B—the first open frontier-level AI model, Meta has released Llama 3.2, which includes small and medium-sized vision LLMs (11B and 90B) and lightweight, text-only models (1B and 3B) that fit onto select edge and mobile devices.

According to Mark Zuckerberg, Founder and CEO, Meta, at the Connect, “This year, Llama has achieved 10x growth and become the standard for responsible innovation. Llama also continues to lead on openness, modifiability, and cost efficiency, and it’s competitive with closed models—even leading in some areas. We believe that openness drives innovation and is the right path forward, which is why we continue to share our research and collaborate with our partners and the developer community.”

The two largest models of the Llama 3.2 collection, 11B and 90B, support image reasoning use cases, such as document-level understanding including charts and graphs, captioning of images, and visual grounding tasks such as directionally pinpointing objects in images based on natural language descriptions.

Meta is making Llama 3.2 models available for download on llama.com and Hugging Face, as well as available for immediate development on their broad ecosystem of partner platforms. Partners are an important part of this work, and for the Llama 3.2 release, they’re also working with on-device partners Arm, MediaTek, and Qualcomm to offer a broad range of services at launch.

Llama 3.2 is poised to reach more people than ever before and enable exciting new use cases. Meta believes that sharing these models with the open source community isn’t enough. Thus, they would ensure that developers also have the tools they need to build with Llama responsibly. Meta has also announced the availability of Llama Stack to the community.