AI marches on. There’s more new and exciting AI coming every day, and today is no exception. Meta Platforms has unveiled Llama 4, the latest iteration of its large language model series. There are two new models: Llama 4 Scout and Llama 4 Maverick. These models are designed to process and integrate multiple data formats, including text, video, images, and audio, enhancing their versatility across various applications.
Llama 4 Scout is a compact model capable of running on a single Nvidia H100 GPU. Meta says it features a 10-million-token context window and has demonstrated superior performance over competitors (such as Google’s Gemma 3 and Mistral 3.1) in various benchmarks. Llama 4 Maverick, a more robust model, delivers performance comparable to OpenAI’s GPT-4o and DeepSeek-V3 in coding and reasoning tasks, while utilizing fewer active parameters.
Meta is also developing Llama 4 Behemoth, an advanced model with 288 billion active parameters and a total of 2 trillion parameters. The company says that Behemoth surpasses models like GPT-4.5 and Claude Sonnet 3.7 on STEM benchmarks. All Llama 4 models employ a “mixture of experts” (MoE) architecture, which activates only the necessary components for a given task, optimizing resource utilization.
Meta plans to integrate these AI models into its Meta AI assistant across platforms including WhatsApp, Messenger, Instagram, and the web. The company wants to reach one billion users by the end of 2025, building on the 600 million users who accessed Meta AI by December 2024.
To support its AI initiatives, Meta says it has committed up to $65 billion in investments for 2025, reflecting the company’s strategic focus on advancing artificial intelligence technologies.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.