Long the default for open models, but Meta's developer site now leads with Muse, so check licences and newer options first.

Kenning's take

What it does

Llama is Meta's family of open-weight language models that developers can download, fine-tune and run on their own hardware or in the cloud. The family includes Llama 2, the Llama 3 series (3.1, 3.2 and 3.3) and Llama 4, which is natively multimodal and uses a mixture-of-experts design, plus safety models such as Llama Guard. Meta's developer site now leads with its newer Muse models, but still hosts Llama guides, licences and downloads.

  • Llama 4 natively multimodal mixture-of-experts models
  • Llama 3.1, 3.2 and 3.3 instruction-tuned models
  • Llama Guard and Prompt Guard safety models
  • Downloads via Hugging Face and Kaggle
  • Guides for running on Mac, Windows, Linux and cloud

Who it suits. Developers and companies who want to run and customise a language model themselves.

Key facts

Category
AI models
Part of
Meta AI
Made by
Meta
Based in
United States
How to get it
free download from Meta, Hugging Face or Kaggle under each version's Llama licence and use policy; also offered by cloud partners
Works on
self-hosted, cloud partners
Free plan
Free weights under the Llama licence for each version

FAQ

Are Llama models free?

Yes, the weights are free to download, under the licence and use policy for each Llama version.

What is the newest Llama?

Meta's Llama page lists Llama 4, a natively multimodal mixture-of-experts family, as its latest Llama release.

Where can I download Llama?

From Meta, Hugging Face or Kaggle.

Facts from the maker's own pages and the sources listed. Checked 2026-10-11.