Large language models (LLMs) are a category of foundation models trained on immense amounts of data. This makes LLMs capable of understanding and generating natural language and other types of content, to perform a wide range of tasks. LLMs can run on a server locally, in order to keep your private data private.