A large language model, or LLM, is an AI model trained on very large amounts of text to understand and generate language. It learns statistical patterns in the text it sees and uses them to produce likely continuations, which lets it answer questions, summarize, translate, write code and more. Many modern chat assistants are built on LLMs.

There is no fixed size at which a language model counts as large, and the term is used loosely. Models vary widely in capability, cost and speed.

LLMs have important limits. They can produce fluent but incorrect statements, often called hallucinations, and they reflect biases and gaps in their training data. They do not know about events after their training cutoff unless given current information, and their outputs can vary from one run to another.

To use one responsibly, verify important claims against reliable sources, avoid sending sensitive data to a service whose handling you have not checked, and test the model on your own tasks before relying on it. Pair it with retrieval or tools when the task needs current or private information.