Meta releases the first two models from multimodal llama 4 suite: llama 4 scout and llama 4 maverick. Maverick is “The Workhorse” of two and above image and understanding of the text for “General Assistant and Cases Using Chat,” The company says a Blog postWhile smaller scout model can face “multi-document summarization, parsing a lot of user activity for personal cennebases.” The company also introduces Llama 4 Behemoth, a future model says “Among the wisest LLMS” in the world Mark Zuckerberg We said we heard about a fourth model, Llama 4 Reason, “next month.”

Both Maverick and Scouts are available to download now from Llama website and Dealing with the faceAnd they were added to Meta Ai, including whatsapp, messenger and Instagram DMS.

Meta

Scout has 17 billion active parameters with 16 experts, said Meta. According to Zuckerberg, “more powerful, native multimodal, and have an industrial leading, almost endless 10 million token contextual context of a GPU.” Maverick on the other side has 17 billion active parameters with 128 experts. The company says competitors are beaten like GPT-4o and Gemini 2.0 in coding, dealing with multikiarical benchats and dealing with Dreepsek v3.1 in reasoning and coding.

Zuckerberg has already called the future model of behemoth, who still trains, “the highest model of the world’s behavior,” with 288 billion active parameters, according to the company. It may not be here yet, but we are likely to hear more about that and the model of reasoning soon; The meta developer conference, Lamacononly a few weeks away.

This article at first shows https engadget: //www..ai/meta-intraltralcal-luma-with-aville-ava2424295.html 1.mlg = ss



Source link