If you have ever wondered what actually happens when you type a question into ChatGPT, Claude, or Gemini, this is the explainer for you. No math, no jargon walls — just a clear mental model you can keep.
What “LLM” actually stands for
LLM means Large Language Model. Break that down: it is large (trained on an enormous amount of text), it works with language, and it is a model — a system that makes predictions. The single job an LLM does is deceptively simple: predict the next word.
Predicting the next word, billions of times
Give the model the phrase “The sky is” and it estimates what word most likely comes next — “blue,” “clear,” “falling.” It picks one, adds it to the sentence, then predicts again. Repeat this thousands of times and you get paragraphs, code, or an essay. Everything an LLM produces is this loop running very fast.
An LLM is not looking anything up. It is generating the most plausible next piece of text, one token at a time.
What is a token?
Models do not read whole words — they read tokens, small chunks of text. “Understanding” might be two tokens: “Understand” and “ing.” This matters because pricing and limits are measured in tokens, not words. A rough rule: 1,000 tokens ≈ 750 English words.
Why it sometimes makes things up
Because the model predicts plausible text rather than true text, it can produce confident, well-formed answers that are simply wrong. This is called a hallucination. It is not lying — it has no concept of truth, only of what usually comes next. That is why you always verify important facts.
The takeaway
An LLM is a very sophisticated autocomplete trained on a huge slice of human writing. Understanding that one idea — next-word prediction — explains most of its strengths and every one of its weaknesses.
