Glossary · Large language model (LLM)
What is a large language model (LLM)?
Definition: A large language model (LLM) is an AI system trained on vast amounts of text to understand instructions and generate written responses.
During training, an LLM learns to predict the next word, or part of a word, in a text. At large scale, that ability allows it to summarise, translate, classify, answer questions or draft documents. It works within a context window: the amount of text it can take into account at once, including the instructions and any documents it is given.
In a business, the model is rarely used on its own. It is the component that interprets language within a wider system: an agent that works with tools, an assistant that answers from internal documentation, or a workflow that extracts data from invoices and contracts. The choice of model depends on the task, the volume, the quality required and where the data must be processed.
An LLM does not look up the answer in a database: it generates the most likely one based on what it has learned. That is why it can be wrong while sounding confident, and why it knows nothing about events after its training or about the company's internal information. Techniques such as RAG supply that information at the moment of answering. It is also worth checking how the provider uses the data sent to it.
Example: an LLM reads each email arriving in the customer service inbox, identifies whether it is a complaint, an order or a question, and extracts the customer number to log it in the CRM.