What is a Large Language Model (LLM)?
Last updated · 1 min read
The one-sentence version
An LLM is a next-word-prediction machine, scaled up until prediction starts to look like conversation.
How it works, without the math
Training an LLM looks like this: feed the model a huge slice of the internet, books and code, and have it practice one simple task — guess the next word, then check the answer, then adjust billions of internal knobs (called parameters) and try again.
Do that at sufficient scale and something surprising happens: to predict text well, the model is forced to learn grammar, facts, style and even reasoning patterns, because those make its guesses better. No one programmed "how to write an email" — the model learned it as a side effect of predicting words.
The second phase is tuning: humans rate the model's answers and it is adjusted to be more helpful and safer. That is why ChatGPT-style assistants feel like an assistant and not an autocomplete.
What it is good at
- Drafting, summarizing, translating, rewriting
- Explaining concepts at the level you ask for
- Working with code and structured text
- Turning a mess of information into a clean first draft
What it is bad at
- Truth. It generates plausible text, not verified facts. Wrong output that looks right is a hallucination.
- Freshness. Its knowledge has a training cutoff unless it is given search or documents — that is what RAG fixes.
- Your private context. It knows the internet, not your company or your files, unless you give it that context.
The practical habit: treat an LLM like a brilliant, fast intern with no consequences for being wrong. Great for drafts; verify before you rely.
Frequently asked questions
- Is ChatGPT an LLM?
- Yes. ChatGPT is a product built on top of large language models made by OpenAI, wrapped in a chat interface with extra safety and search features.
- Does an LLM understand what it writes?
- It processes language patterns at a scale that produces useful, often correct output, but it has no human-style understanding or awareness. Treat its output as plausible text that needs checking, not as verified truth.
- Why does an LLM make things up?
- Because it is built to produce the most plausible next words, not to look facts up. When it lacks knowledge it will still produce confident-sounding text — that is called hallucination.
New to AI? Start with the guides or browse the plain-language glossary.