Meta unveils Llama 4 AI models, claiming top advancements

In early April 2025, Meta announced the release of its latest AI models under the Llama 4 family, which includes Scout, Maverick, and Behemoth. The company claims that these multimodal models are the most advanced in its open-source AI ecosystem. Scout and Maverick have been noted for their ability to handle a variety of data types and were trained on substantial amounts of unlabelled text, images, and video data. Meta boasts that Scout outperforms competitors like Google’s Gemma 3, while Maverick provides comparable results to DeepSeek V3 in reasoning and coding tasks. The Llama 4 models also exhibit a remarkable capacity for managing extensive input data, with Scout boasting a 10 million token context window, enabling it to process five million words and thousands of images. This capability allows Meta to potentially create accurate digital representations of its users based on their social media activity. However, the release of these models comes amid ongoing scrutiny regarding their training datasets, which are alleged to contain copyrighted material used without permission, raising concerns about data privacy and ethical implications in the AI landscape. The introduction of Llama 4 coincides with a growing AI race that has seen various companies, including Chinese startup DeepSeek, gain traction with their models. DeepSeek’s recent success has seemingly influenced Meta’s AI development trajectory. Meta claims that both Scout and Maverick deliver superior performance compared to existing AI models. In light of these advancements, Behemoth is described as Meta’s most powerful AI model yet, intended to further guide the development of future AI models as it undergoes training of its own. However, it is important to note that users and companies in the European Union face restrictions in accessing the Llama models due to stringent regulations on how AI models may be trained utilizing EU data. This has prompted Meta to describe the regulations as unpredictable, signaling potential barriers to the widespread adoption of their latest AI advancements in certain markets. The AI conversation continues to evolve, especially as companies like Meta navigate the complexities of ethical AI use and content generation.