Meta has launched the primary two fashions from its multimodal Llama 4 suite: Llama 4 Scout and Llama 4 Maverick. Maverick is “the workhorse” of the 2 and excels at picture and textual content understanding for “common assistant and chat use circumstances,” the corporate mentioned in a weblog put up, whereas the smaller mannequin Scout may deal with issues like “multi-document summarization, parsing in depth person exercise for personalised duties, and reasoning over huge codebases.” The corporate additionally launched Llama 4 Behemoth, an upcoming mannequin it says is “among the many world’s smartest LLMs” — and CEO Mark Zuckerberg mentioned we’ll be listening to a couple of fourth mannequin, Llama 4 Reasoning, “within the subsequent month.”
Each Maverick and Scout can be found to obtain now from the Llama web site and Hugging Face, and so they’ve been added to Meta AI, together with for WhatsApp, Messenger and Instagram DMs.
Scout has 17 billion lively parameters with 16 consultants, Meta says. In keeping with Zuckerberg, “It’s extraordinarily quick, natively multimodal, and has an trade main, almost infinite 10 million token context size, and it’s designed to run on a single GPU.” Maverick alternatively has 17 billion lively parameters with 128 consultants. The corporate says it beats opponents like GPT-4o and Gemini 2.0 on coding, reasoning, multilingual, long-context and picture benchmarks, and stacks up in opposition to DeepSeek v3.1 on reasoning and coding.
Zuckerberg is already calling the upcoming Behemoth mannequin, which remains to be coaching, “the very best performing base mannequin on the earth,” with 288 billion lively parameters, in accordance with the corporate. It will not be right here but, nevertheless it’s possible we’ll be listening to much more about that and the Reasoning mannequin quickly; Meta’s huge AI developer convention, LlamaCon, is just some weeks away.
