Meta has unveiled the latest enhancements to its Llama series, introducing three new AI models: Scout, Maverick, and Behemoth.
These additions are poised to strengthen Meta’s standing within the AI industry. The models are engineered to handle vast amounts of data, having been trained on a diverse collection of unlabeled text, images, and audio.
Both Scout and Maverick are now available via Llama’s official website and the Hugging Face platform. Scout is a multimodal model capable of processing various types of data—text, images, and audio—simultaneously. This versatility allows it to excel in tasks such as document summarization and complex reasoning across large datasets. Meta reports that Scout can quickly analyze texts exceeding five million words, making it highly effective for examining intricate research papers, meetings, and similar materials.
Conversely, Maverick acts as a general-purpose assistant proficient in understanding text and images. While it possesses a greater number of parameters and a more expansive knowledge base than Scout, it does exhibit limitations in contextual memory. This indicates that, even though it can tackle a broader range of tasks, it might struggle with retaining lengthy conversations.
Behemoth, the most advanced and largest model in Meta’s lineup, does not yet have a release date as the company continues to refine it. However, it is anticipated that Behemoth will merge the strengths of Scout and Maverick, synthesizing extensive knowledge with improved contextual comprehension. Internally, Meta asserts that Behemoth has surpassed several leading AI models, such as OpenAI’s GPT-4.5, Claude 3.7 Sonnet, and Google Gemini 2.0 Pro, in internal assessments focused on STEM-related tasks, particularly in advanced problem-solving.
Meta intends to integrate the new Llama 4 models into its platforms: WhatsApp, Facebook Messenger, and Instagram. The new AIs will roll out in 40 countries, although EU regulations will prevent European users from accessing them.
Image Source: Cristian Valderas / Shutterstock
