On February 24, 2023, Meta announced LLaMA in a post titled “Introducing LLaMA: A foundational, 65-billion-parameter large language model.” Meta described publicly releasing LLaMA (Large Language Model Meta AI) as a state-of-the-art foundational large language model intended to help researchers advance their work, and noted the model came in several sizes: 7 billion, 13 billion, 33 billion, and 65 billion parameters.
The accompanying paper, “LLaMA: Open and Efficient Foundation Language Models” by Hugo Touvron, Guillaume Lample, and a team at Meta, was submitted on February 27, 2023. It argues that smaller models trained on more data can match or exceed much larger models, making capable language models more efficient to run.
LLaMA became a catalyst for the open-weights movement. By putting capable foundation models into the hands of the research and developer community, it sparked a wave of derivative models and fine-tuned variants, giving businesses an alternative to closed commercial systems and intensifying competition across the field.