A large language model (LLM) is an AI model trained on vast amounts of text to understand and generate human-like language, powering tasks such as answering questions, summarising, and writing code.
How it works
An LLM learns statistical patterns of language by training on huge text corpora, then predicts the most likely next piece of text given what came before. Modern LLMs are built on the transformer architecture, which lets them weigh the relationships between words across long passages to produce coherent, context-aware responses.
Why it matters for enterprise AI
LLMs are the engine behind most generative and agentic AI. On their own, though, they answer from training data and can be out of date or wrong for a specific business — which is why enterprises pair them with their own data (via techniques like retrieval-augmented generation) and governance before trusting them in production.
