What is a large language model (LLM)?

In plain English

A large language model, or LLM, is the technology behind AI chatbots and writing assistants. It is a very large neural network trained on an enormous quantity of text — books, websites, code, transcripts — with one deceptively simple objective: given some text, predict what comes next.

Do that at sufficient scale and something surprising emerges. To predict text well, the model has to absorb grammar, facts, writing styles, the structure of arguments and a good deal of everyday reasoning. The result is a system you can converse with, ask to summarise a document, or instruct to write in a particular tone.

Models work with “tokens” — chunks of text roughly the size of a short word or word fragment — and hold a limited amount of conversation in view at once, known as the context window.

Why it matters

LLMs made AI usable through plain language rather than code, which is why they spread through offices, classrooms and customer service so fast. They are also general-purpose: the same model can draft an email, explain a contract clause and translate a paragraph, without being rebuilt for each job.

An example

You paste a twelve-page supplier agreement into a chatbot and ask for the key obligations and termination terms in bullet points. It returns a readable summary in moments — which you then check against the actual document before relying on it.

What to watch out for

An LLM is predicting plausible text, not looking up truth. It can state wrong facts confidently, invent sources, and misread what you asked. Its knowledge comes from training text gathered up to a point in time, so it may be out of date unless connected to a search tool. And anything you type may be stored or reviewed, so treat confidential information carefully.