Meta released the first two models from her multimodal suite Llama 4: Llama 4 Scout and Llama 4 Maverick. Maverick is “the battle horse” of the two and excels in the image and understanding of the text for “the cases of general use of assistant and cat,” said society in a blogWhile the smallest scout could tackle things like “a multi-documentary summary, analyze an in-depth user activity for personalized tasks and reason on large code bases”. The company also introduced the giant of Llama 4, an upcoming model which, according to him, is “among the most intelligent LLM in the world” – and CEOs Mark Zuckerberg We have said that we will hear about a fourth model, the reasoning of Llama 4, “next month”.
Maverick and Scout are available for download now from the Lama website and FaceAnd they were added to Meta Ai, including for WhatsApp, Messenger and Instagram DMS.
Scout has 17 billion active parameters with 16 experts, says Meta. According to Zuckerberg, “it is extremely fast, natively multimodal, and has a context of 10 million tokens of 10 million 10 million industry, and it is designed to operate on a single GPU.” Maverick, on the other hand, has 17 billion active parameters with 128 experts. The company says that it beats competitors like GPT-4O and Gemini 2.0 on coding, reasoning, multilingual references, long-term and image, and accumulates against Deepseek v3.1 on reasoning and coding.
Zuckerberg already calls the next giant model, which is still in training, “the most efficient basic model in the world”, with 288 billion active parameters, according to the company. It may not be there yet, but it is likely that we hear much more about this and on the model of reasoning soon; The great conference of Meta AI developers, Llamacon, is a few weeks old.
This article originally appeared on engadget to
