Large language model (LLM)
In shortA large language model (LLM) is an AI system trained on huge amounts of text to predict the next token, which lets it write, summarise and answer.
In plain words
A large language model, or LLM, is a program that has learned patterns in very large amounts of text. Its core task is simple: given the text so far, predict the next token. It repeats this step again and again, and the chain of predicted tokens becomes a sentence, an email or a block of code.
Anthropic's glossary describes the first phase as pretraining: the model learns to predict the next word from a large unlabeled text collection. Such a raw model is not good at following instructions. A second phase, fine-tuning and training with human feedback, turns it into a helpful assistant. Most modern LLMs use the transformer design introduced in the 2017 paper "Attention Is All You Need". The size of a model is often given as its number of parameters.
A chatbot is not the same as an LLM. A product like a chat app wraps an LLM with a system prompt, a memory, tools and a user interface.
Why it matters
- Strengths: drafting, rewriting, summarising, translating, explaining and writing code.
- Limits: An LLM generates likely text. It can state false things with confidence (hallucination), it knows nothing past its knowledge cutoff unless it gets tools, and the same prompt can give different answers.
- Choice: Models differ in quality, speed, context window and price. Choose a model for the task, not for the name.
- Data: What you type may leave your device. Read what never to paste into a chatbot.
Example
You paste three paragraphs of meeting notes and write: "Summarise in three bullets for a manager." The model turns your text into tokens, predicts the first token of its answer, adds it to the text, and predicts the next one. After a few dozen steps you have three bullets. No database lookup took place; the wording came from patterns the model learned.
Common misconceptions
- "An LLM searches the web." Not by itself. Search is a tool that an app can add.
- "It stores a copy of its training texts." It stores learned weights, not a library of documents, though it can sometimes repeat text it has seen often.
- "LLM and ChatGPT mean the same." ChatGPT is one product built on LLMs. Many products and many model families exist.
Questions and answers
What does LLM stand for?
Large language model. It is an AI model trained on very large amounts of text to predict the next token.
How is an LLM different from a chatbot?
The LLM is the model. A chatbot is an app around it, with a system prompt, tools, memory and an interface.
Does an LLM understand what it writes?
It predicts text from learned patterns. Whether this counts as understanding is debated, so check important facts yourself.
Read next
- Token
A token is a small chunk of text, often a word piece, that an AI model reads, writes and bills.
- Parameters
Parameters are the numbers, also called weights, that a model adjusts during training and then uses to answer; their count is a rough measure of size.
- Hallucination
A hallucination is a fluent, confident AI answer that is false, invented or not backed by the sources it was given.
- Fine-tuning
Fine-tuning trains an existing model further on your own examples, so it follows a style, format or task more reliably.
- How to choose an AI model
Match the model to the job: small for volume, mid for daily work, top or reasoning for hard problems and local for private data, with a decision tree and a test.