Meta has unveiled its latest open-source AI model, Llama4, marking another significant advancement in the field of artificial intelligence. Llama4 comes in two versions, named Scout and Maverick, designed to enhance AI model capabilities and performance. Meta states that Llama4 is a multimodal large language model capable of processing various data types, including text, images, videos, and audio, and seamlessly converting between these formats.
Notably, the Llama4 series is the first to employ a Mixture-of-Experts (MoE) architecture. This innovative design significantly improves the model's training and response efficiency. By dividing the model into multiple "expert" sub-models focused on specific tasks, Llama4 is faster and more accurate when handling complex queries.
According to Meta, Scout and Maverick are currently the company's most advanced models. Scout boasts 17 billion active parameters, equipped with 16 "expert" models, resulting in a total of 109 billion parameters. Maverick also has 17 billion active parameters but features 128 "expert" models, leading to a staggering 400 billion total parameters. Furthermore, Meta is developing a more powerful model called Behemoth, described as the "teacher" for new models, hinting at future innovations.
These models, trained on a massive amount of unlabeled data, demonstrate exceptional visual understanding capabilities. Scout and Maverick are now publicly available on Llama.com and Meta's partner platforms, and have already been integrated into social media platforms like WhatsApp, Messenger, and Instagram Direct.
Meta CEO Mark Zuckerberg stated at the launch, "Our goal is to build world-leading AI and make it open-source so more people can benefit." He emphasized that the launch of Llama4 signifies the rise of open-source AI, poised to reshape the industry landscape.
Meta plans to invest up to $65 billion in expanding its AI infrastructure over the next year, despite investor caution regarding the actual returns of large tech companies. These actions will undoubtedly enhance Meta's competitiveness in the AI field and drive technological innovation across the industry.