In short
A large language model (LLM) is an AI model trained on very large text corpora that can understand natural language and generate text.
LLMs are used for Q&A, summarisation, translation, document classification and code generation. In enterprise use, the model is grounded in company documents and system data (RAG) so it answers in the right context.
When confidential data is involved, hosting the model on-premise, enforcing access rights and making answers verifiable with cited sources are essential.