LLaMA

LLaMA, short for Large Language Model Meta AI, is a family of transformer-based large language models developed by Meta AI. Released in 2023, the LLaMA line was positioned as an efficient and accessible set of foundation models intended for research and commercial use. Like other autoregressive language models, LLaMA generates text by predicting subsequent tokens from a given prompt, and it can be adapted for tasks such as completion, summarization, translation, and question answering.

The LLaMA-1 generation offered models of approximately 7B, 13B, 30B, and 65B parameters, trained on a mixture of publicly available text data and licensed sources using a high-performance transformer architecture. In 2023, Meta introduced LLaMA-2, a follow-on series available in sizes including 7B, 13B, and 70B parameters, with updates emphasizing improved instruction-following and safety features and broader licensing terms to support research and commercial use.

Access to LLaMA weights initially followed a restricted invitation model, limiting usage to researchers and organizations under specific terms. With LLaMA-2, Meta released the models under licenses designed to broaden access for both research and commercial applications, though users must comply with the license terms and responsible-use guidelines.

Since its release, LLaMA has been used in academia and industry for a range of natural language processing tasks, including benchmarking, prototype development, and as a basis for instruction-tuned variants by researchers and practitioners. The models contribute to ongoing discussions about open access to large language models and the balance between performance, safety, and licensing.