Meta announced the latest large-scale language model, Llama 4, designed as a native multi-modality AI model that can process both text and images.
Llama 4 is a multimodal large model capable of processing text, video, images, and audio, and converting content between these formats.
It is the first in the Llama series to utilize a mixture of experts (MoE) architecture, enhancing efficiency in training and addressing user queries by dividing the model into specialized 'expert' sub-models.
Mark Zuckerberg emphasized the aim of creating the leading AI, open-sourcing it, and ensuring accessibility for all. Llama 4 is available in three main variations.