Llama

Executive Definition: Llama — A family of open-weight large language models developed and released by Meta.

The Llama (Large Language Model Meta AI) family includes several generations of models ranging in size from small versions suitable for local deployment to massive, state-of-the-art models competitive with proprietary APIs. Because their weights are publicly accessible, Llama models serve as a foundational building block for the open AI ecosystem, enabling researchers and developers to fine-tune, modify, and deploy powerful language models on their own infrastructure.

Frequently Asked Questions

Are Llama models truly open-source?

They are open-weight, meaning the model weights are available for download and use. However, they are released under a custom Meta license with specific restrictions, so they do not meet the strict OSI definition of open-source.

What does the ‘B’ mean in Llama models (e.g., 8B, 70B)?

The ‘B’ stands for billions of parameters. A parameter is a learned weight in the neural network. Larger parameter counts generally correlate with better reasoning and capability but require more computational power to run.

How can developers use Llama models?

Developers can run them locally using tools like Ollama or vLLM, host them on cloud providers, or access them via API endpoints from managed services like Groq, Together AI, or AWS Bedrock.