large language model (LLM)

A language model built from a Transformer with billions of parameters, pretrained on a huge amount of text.

A large language model (LLM) is a language model whose parameter count and training-corpus size are large enough that the model develops broad, general-purpose capabilities -- following instructions, answering questions on topics never explicitly labeled during training, writing code -- rather than performing only the single task it was narrowly trained for.

Examples

A chat assistant is an LLM wrapped in a conversation format and a serving system.