Meta has released Llama 4 Maverick and Scout, two new open-source language models that are now available for download on Hugging Face. These models represent the latest iteration of Meta's Llama family, designed to offer improved performance and efficiency for a wide range of natural language processing tasks.
Llama 4 Maverick and Scout come with distinct capabilities. Maverick is the larger model, optimized for complex reasoning and creative generation, while Scout is a smaller, more efficient variant suitable for deployment on resource-constrained devices. Both models are released under the Llama 4 Community License, which permits research and commercial use.
The models were trained on a diverse dataset and incorporate advanced techniques such as reinforcement learning from human feedback (RLHF) to align with user expectations. Early benchmarks suggest that Llama 4 models achieve competitive results against other open-source alternatives like Mistral and Gemma, while maintaining a strong commitment to safety and responsibility.
Developers and researchers can access the models directly from Hugging Face, where they are available for download along with documentation and example code. The release is expected to accelerate innovation in the AI community by providing powerful tools that can be customized and integrated into various applications.