LLM (large language model)
Also called large language model, foundation model.
An AI model trained on very large amounts of text that predicts the next piece of text, which lets it read, write and reason in language.
Last updated
What it is.
An LLM is trained to predict the next token in a sequence. Done at very large scale, that simple goal produces a model that can answer questions, summarise, translate, write code and follow instructions. Many current models also accept images, audio or files.
Apps use an LLM through an API or a self-hosted model. The model knows only what was in its training data and what you send in the context window, so most products add their own data with RAG and their own actions with tool calling.
Why it matters
LLMs make it practical to automate work that involves reading and writing: support, document handling, search, drafting and data extraction. They are also probabilistic, so the same input can give different outputs and they can be confidently wrong. Building with them means designing for checks, not just prompts.
Common mistakes.
- 01
Treating the model as a database of facts. Ground answers in your own data.
- 02
Shipping without evals. Without them you cannot tell if a prompt or model change made things better or worse.
- 03
Picking the biggest model by default. A smaller model is often good enough and cheaper per call.
Work it out with a tool.
Related terms.
- Token (LLM)The unit an AI language model reads and writes, and the unit most LLM APIs bill by.
- RAG (retrieval-augmented generation)A pattern that finds relevant passages from your own data and gives them to a language model along with the question, so it answers from your content.
- Hallucination (AI)When an AI model produces information that is false or made up, presented as if it were true.
- Fine-tuningTraining an existing model further on your own examples so it behaves the way you want, such as following a format, tone or task more reliably.
Guides and pages
Browse the full glossary or read answers to common questions.
