Meta has announced Llama 4, its newest collection of AI models that now power the Meta AI assistant on the web and in WhatsApp, Messenger, and Instagram. The two new models, also available to download from Meta or Hugging Face, are Llama 4 Scout — a small model capable of “fitting in a single Nvidia H100 GPU” — and Llama 4 Maverick, which is more akin to GPT-4o and Gemini 2.0 Flash. Meta says it’s still in the process of training Llama 4 Behemoth, which Meta CEO Mark Zuckerberg says is “the highest performing base model in the world.”
According to Meta, Llama 4 Scout has a 10-million-token context window — the working memory of an AI model — and beats Google’s Gemma 3 and Gemini 2.0 Flash-Lite models, as well as the open-source Mistral 3.1, “across a broad range of widely reported benchmarks,” while still “fitting in a single Nvidia H100 GPU.” Meta makes similar claims about its larger Maverick model’s performance versus OpenAI’s GPT-4o and Google’s Gemini 2.0 Flash, and says its results are comparable to DeepSeek-V3 in coding and reasoning tasks using “less than half the active parameters.”
Meanwhile, Llama 4 Behemoth has 288 billion active parameters with 2 trillion parameters in total. While it hasn’t been released yet, Meta says Behemoth can outperform its competitors (in this case GPT-4.5 and Claude Sonnet 3.7) “on several STEM benchmarks.”
For Llama 4, Meta says it switched to a “mixture of experts” (MoE) architecture, an approach that conserves resources by using only the parts of a model that are needed for a given task. The company plans to discuss future plans for AI models and products at its LlamaCon conference, which is taking place on April 29th.
As with its past models, Meta calls the Llama 4 collection “open-source,” although Llama has been criticized for its license restrictions. For instance, the Llama 4 license requires commercial entities with more than 700 million monthly active users to request permission from Meta before using its models, which the Open Source Initiative wrote in 2023 takes it “out of the category of ‘Open Source.’”