Llama

Llama is Meta's family of open-source language models and the foundation of the open-weight AI ecosystem. Llama 4 introduced mixture-of-experts with Scout (109B total, 17B active parameters) and Maverick (400B total, 17B active), delivering strong performance with efficient inference. The models are free for commercial use, which has made them the default choice for companies that need to self-host or fine-tune. The ecosystem around Llama is massive, with support in every major inference framework, fine-tuning toolkit, and deployment platform. You can run smaller variants locally through Ollama, or deploy the full models on your own GPU infrastructure. For developers who need full control over their model stack without licensing restrictions, Llama is the starting point.