Llama (model)

← All terms · Local models and inference

Also called Llama 3, Llama 4

A dominant family of open-weight foundational models developed by Meta, serving as the backbone for much of the open-source AI ecosystem.

What it is

Meta's Llama models made it possible for anyone to run frontier-quality models locally by releasing very capable, large models (up to 400B+ parameters) for free public use. Because of their immense popularity, almost all local inference tools, quantization formats, and developer frameworks optimize for Llama architectures first.

When you would use it

You use Llama when you want the most widely supported, community-tested open-weight model available, ensuring maximum compatibility with third-party tools.

Common operations

• Fine-tuning a base Llama model for a highly specific enterprise task.
• Running a 8B parameter Llama model locally on a laptop for fast text processing.

Related terms

• Mistral
• Qwen
• Quantization

Where this is taught

No learning path uses this term yet. Browse the Learning Atlas for guided sequences through related ideas.