A large language model (LLM) is an AI model consisting of a neural network with typically billions of weights or more. It is trained on unlabelled text using self-supervised learning, as these models are far too large to make labeled input data viable. BERT, and most recently GPT3 and GPT4 are examples of LLMs.