What is a large language model (LLM)?

A large language model (LLM) is an AI model trained on very large amounts of text so that it can understand and produce human language. A large language model works by predicting the most likely next piece of text, which lets it answer questions, summarise documents, draft content and write code.

Why it matters for your business

Large language models are the engine inside most of the AI tools a business now meets: chat assistants, meeting summarisers and AI agents. They work with ordinary language, such as emails, contracts and reports, which older software could not read. Knowing roughly how they work helps a leader judge what to trust them with.

How it works

The model is trained on a huge body of text and learns the patterns of language. Text is broken into small units called tokens: words or parts of words. Given a request, it predicts the most likely next token, again and again, until the answer is complete.

The model does not look facts up. It produces what is statistically likely.

Costs, limits and when not to use it

Hallucination: a model can state something false in a confident way. Anything that matters needs checking. No knowledge of your business: it knows only what was in its training data, plus whatever is supplied with the request. Bias: it can repeat biases in the text it learned from. Confidentiality: check a provider's terms before staff enter confidential information.

For exact arithmetic, or decisions that must come out the same every time, conventional software is the better tool.

Large language model vs generative AI

Generative AI is the broad category of AI that creates content, including images, audio and video. A large language model is the kind that works with text.