Meta has launched its Llama 4 series with two models, Scout and Maverick, designed for advanced language and image processing tasks. Scout features 17 billion parameters for efficient document summarization, while Maverick is built for general AI tasks with enhanced capabilities. Both models contribute to the upcoming Llama 4 Behemoth supermodel, which is expected to be immensely powerful. They are available for download through various platforms under a commercial and research license.
Meta has unveiled its latest series of artificial intelligence models, Llama 4, featuring two primary models: Scout and Maverick. Both models are designed to address complex tasks involving language and image processing and are notable for their advanced capabilities compared to their rivals in the AI landscape.
Llama 4 Scout, the more compact of the two, boasts 17 billion active parameters and 16 experts. This nimble model excels in summarizing numerous documents, analyzing extensive user data, and drawing insights from code. Meta claims that Scout operates swiftly, accommodates a context of up to 10 million tokens, and can effectively run on a single GPU card.
On the other hand, Llama 4 Maverick serves as the primary “workhorse,” also featuring 17 billion active parameters but with 128 experts onboard. Tailored for general AI applications, it shines in code evaluations, long-context tasks, multilingual interactions, and image recognition. Meta highlights Maverick’s advantages over competitors like GPT-4o and Gemini 2.0, particularly in inference and programming capabilities.
Both models contribute to the development of the Llama 4 Behemoth supermodel, which is still undergoing training and is anticipated to have 288 billion active parameters. This supermodel aims to be among the most formidable foundation models globally. Additionally, Llama 4 Reasoning, another soon-to-be-revealed model, will focus on advanced inference skills.
Meta’s new models, available for download through its website and platforms like Hugging Face, Azure, AWS, and Google Cloud, come under a commercial and research license, marking a significant stride in the field of AI deployment.
Meta’s introduction of the Llama 4 series, featuring Scout and Maverick, represents a noteworthy advancement in AI technology. With remarkable capabilities in language and image processing, these models cater to an array of applications while positioning themselves favorably against their competitors. The anticipation surrounding the eventual release of the Llama 4 Behemoth supermodel only adds to the excitement, pushing the boundaries of what artificial intelligence can achieve.
Original Source: www.jpost.com
