A large language model, or LLM, is a computer program trained to process and output natural language through text. You can view the model as learned knowledge from a library of resources (database) on the topic of language. Some training material sets are small while others are large. Through the training, the model can predict a string of text it should produce in response to a prompt, making it appear as though it were aware and thinking.
The language model can generate human language through text using algorithms to learn pattern recognition and relationships between words from the training material. This system of training is called machine learning. Typically, a user can interact with the LLM through some interface.