What is Llama (Meta's open LLM)?

Llama is Meta's family of large language models, and its defining trait is that Meta publishes the actual model weights for free download instead of locking the model behind an app or API like ChatGPT or Claude.

What does "open-weight" mean, and why does it matter?

It means any developer, researcher, or company can pull Llama onto their own servers, inspect it, and fine-tune it for their own use case, without paying per-query fees to Meta or routing sensitive data through someone else's cloud. This "open-weight" release strategy — not fully open-source, since Meta's license carries some usage restrictions — is what set off much of the open-model boom over the past few years. Startups building products, researchers running experiments, and companies with privacy or cost concerns all lean on Llama because it gives them a capable model they can own and control, instead of depending entirely on a single vendor's pricing and rules.

What is the current version of Llama?

The current generation is Llama 4, released as two main versions, Scout and Maverick, both still downloadable from Meta and Hugging Face. Meta's license lets almost any company use Llama commercially, though a couple of conditions apply:

• Businesses with more than 700 million monthly users need a separate deal with Meta.
• EU users face some added restrictions.

Is Meta still committed to open Llama?

Worth knowing: in 2026, Meta's newest, most advanced AI work shifted toward a closed, API-only model built by its Superintelligence Labs team, separate from the open Llama line. That doesn't erase what Llama did — it's still one of the main reasons a genuinely open alternative to closed chatbots exists at all.