Llama 4

Llama 4 is an open-source large language model series released by Meta and the fourth generation of the Llama model family. The series adopts advanced architectures including Mixture of Experts (MoE) and features multimodal understanding and long-context processing capabilities. Model weights are released to researchers and developers, supporting local deployment and fine-tuning across multiple platforms for tasks such as dialogue, reasoning, and code generation.