What is Large Language Model (LLM)?
A model trained on huge amounts of text that can generate and understand language.
A large language model is trained on vast amounts of text to predict the next word, which lets it generate and understand human language.
LLMs power chatbots, writing assistants, coding copilots, and most text-based AI tools.
A large language model learns language by being trained on enormous amounts of text with a deceptively simple objective: predict the next word (technically, the next token) given everything before it. Repeated across trillions of words, this forces the model to absorb grammar, facts, reasoning patterns and writing styles, because predicting the next word well requires understanding context. The "large" refers to both the training data and the number of parameters — the internal values it learns — which now run into the hundreds of billions.
Once trained, an LLM generates text one token at a time, each time predicting the most likely continuation and feeding its own output back in. That single mechanism — next-token prediction — is what underlies chatting, drafting, summarising, translating and coding. It also explains the model's limits: it predicts plausible text rather than retrieving verified facts, which is why it can produce fluent answers that are wrong, and why what it "knows" is frozen at its training cutoff unless connected to live data.
Why it matters
LLMs are the engine behind almost every text-based AI tool you use — chatbots, writing assistants, coding copilots, support bots and research tools all sit on top of one. Understanding that they predict plausible text rather than look up facts is the single most useful thing to know, because it explains both their power (fluent, flexible language) and their failure modes (confident errors, stale knowledge).
A concrete example
When you type "summarise this email in two sentences," the LLM doesn't search a database of summaries. It predicts, token by token, the text most likely to follow your instruction given the email — producing a fresh summary that didn't exist before, the same way it would draft a poem or explain code.
Where you’ll meet it in AI tools
Every general AI assistant (ChatGPT, Claude, Gemini) is an interface to an LLM, and most AI writing, coding, support and research tools are built on one — sometimes the same underlying model, differentiated by the product around it. When tools advertise a "context window," "reasoning" mode or "knowledge cutoff," they're describing properties of the LLM underneath.
Large Language Model (LLM): FAQ
Is an LLM the same as AI?
No — AI is the broad field, and an LLM is one specific kind of AI model that works with language. Many AI tools (image generators, recommendation systems, computer vision) don't use LLMs at all. But because LLMs power the chatbots and writing tools most people interact with, they're what many now picture when they think of "AI." An LLM is a powerful example of AI, not a synonym for it.
Why do LLMs make things up?
Because they generate plausible text rather than retrieve verified facts. An LLM predicts the most likely next words given your prompt, and when it lacks the right information it will still produce a fluent, confident-sounding continuation — which can be wrong. This is called hallucination. It's a fundamental property of how they work, which is why important answers should be verified and why techniques like retrieval-augmented generation (RAG) connect models to real sources.
Do all AI chatbots use the same LLM?
No. Different assistants use different underlying models — some build their own, some use a provider's — and they differ in reasoning ability, context window, training data and tuning. Many specialised tools also build on a general provider's model via an API. That's why the same prompt can get noticeably different answers from different chatbots, and why it's worth testing your real task across a few.
Related terms
Find language models tools
Browse hand-reviewed AI tools — compare on pricing, features and real ratings.