Llama Glossary

*Meta AI*'s open-weights language model family, first released February 2023. The most-downloaded open-source LLM line and the foundation under most of the post-2023 open-weights ecosystem.

Llama (originally styled LLaMA — Large Language Model Meta AI) is the open-weights large language model family released by Meta AI. The first generation (Llama 1) was released February 2023 under a research-only licence; Llama 2 (July 2023) shifted to a more permissive licence; Llama 3 (April 2024) and Llama 4 (April 2025) have continued the open-weights posture with progressively larger parameter scales and improved capabilities. As of 2026, Llama is the most-downloaded open-weights LLM family in the world by a wide margin and the practical foundation under most of the post-2023 open-weights ecosystem.

For this Dictionary, Llama matters at two levels. First, technically: Llama base models are the canonical foundation that other open-weights labs (most notably Nous Research with Hermes) build fine-tuned variants on top of. Second, strategically: Meta’s sustained open-weights posture is the single most important reason the Sovereign Compute argument is viable at all. Without Llama (or a credible replacement) the gap between frontier closed-weights models and the best open-weights options would be much wider than it currently is, and the Dictionary’s Open-Weights Inversion thread would not have an empirical foundation.

The licensing remains open-ish rather than fully open — Meta restricts commercial use by very large companies (>700M MAUs) and requires a license-acceptance step — but for nearly all individual operators the Dictionary serves, the licence is functionally equivalent to fully open.

See also

Return to Dictionary All Entries (A–Z) For Students Other Writing