What Is an LLM (Large Language Model)?

An LLM — short for large language model — is the kind of AI that powers chatbots like ChatGPT and Claude. In plain terms, it’s a system trained on enormous amounts of text that has learned to do one thing astonishingly well: predict the next word. Everything these tools can do — writing, answering, summarizing, coding — grows out of that single skill, scaled up to a staggering degree. This is the engine under the hood; if you want to see how it powers a chatbot specifically, our explainer on how AI chatbots work builds on the same idea.

What “LLM” actually stands for

Let’s unpack the three words:

Large — the model is trained on a vast slice of text (books, websites, code) and has a huge number of internal settings, called parameters, often in the billions.
Language — its job is working with human language: reading it and generating it.
Model — in AI, a “model” is a mathematical system that learns patterns from data and uses them to make predictions.

Put together, an LLM is a very large pattern-learning system specialized in language.

How LLMs work, in one idea

Here’s the core: an LLM is really, really good autocomplete. Think of your phone suggesting the next word, scaled up almost unimaginably. Given the text so far, the model calculates the most likely next chunk of text, adds it, then repeats — chunk by chunk — until it has built a full response.

The surprising part is what emerges from doing this at scale. To predict the next word well across trillions of words, the model has to absorb grammar, facts, reasoning patterns, coding conventions and writing styles. None of that is programmed in by hand — it’s a side effect of getting prediction right. That’s why an LLM can explain a concept or write a poem despite “only” predicting text.

Tokens, parameters and the context window

Three terms are worth knowing because you’ll see them everywhere:

Tokens are the chunks of text an LLM reads and writes — roughly a word or part of a word. Models think in tokens, not letters.
Parameters are the model’s internal dials (billions of them) that get tuned during training. Loosely, more parameters can mean more capability, though it’s not the whole story.
The context window is how much text the model can consider at once — your conversation so far plus the latest message. Exceed it and the earliest parts start to fall out of view.

How an LLM is trained

Building one happens in stages:

Pretraining. The model reads a massive amount of text and practices predicting the next token over and over, adjusting its parameters until it’s good at it. This is where most of the raw knowledge comes from.
Fine-tuning. It’s then trained on higher-quality examples to behave like a helpful assistant rather than a raw text predictor.
Human feedback. People rate its answers, and the model is nudged toward responses humans prefer — more helpful, honest and safe.

Underneath it all is an architecture called the transformer, whose key trick — attention — lets the model weigh which earlier words matter most when predicting the next one. That’s what keeps it on track across a long passage, and it’s the breakthrough that made today’s LLMs possible.

Why LLMs sometimes get things wrong

This is the most important thing to understand: an LLM predicts plausible text; it doesn’t look up verified facts. When it doesn’t know something, it can generate a confident, well-formed answer that’s simply wrong — a so-called hallucination. It isn’t lying; it’s doing exactly what it was built to do (produce likely-sounding text), which isn’t the same as being correct.

So the golden rule is: use an LLM to draft and explain, but verify anything that matters.

Where you’ll meet LLMs

You’re already using them. LLMs power ChatGPT, Claude, Google’s Gemini and others — and choosing between them is really about which model and which features fit your work (see our Claude vs ChatGPT breakdown). Connect an LLM to tools like web search or code execution and let it act in a loop, and it becomes an AI agent that completes tasks rather than just answering. And if you want software built on these models for a specific job, see the best AI writing tools.

FAQ

What does LLM stand for?

LLM stands for “large language model.” It’s a type of AI trained on huge amounts of text that generates language by predicting the next chunk of text, one token at a time.

How does a large language model work?

It works like extremely powerful autocomplete: given the text so far, it predicts the most likely next token, adds it and repeats. Doing this well at massive scale lets it write, answer and reason, even though it’s fundamentally predicting text.

What’s the difference between an LLM and a chatbot?

The LLM is the underlying model — the engine. A chatbot is the app wrapped around it, with a chat interface and extra features. ChatGPT and Claude are chatbots powered by LLMs.

Why do LLMs make mistakes?

Because they generate statistically likely text rather than looking up verified facts. When unsure, they can produce confident but incorrect answers, called hallucinations — so always verify anything important.

AI models and their capabilities change quickly. This explainer covers the durable fundamentals and is reviewed periodically.