Large language model (LLM)

An LLM learns by reading enormous amounts of text and repeatedly predicting the next token, a word or piece of a word. Doing this billions of times teaches it grammar, facts, styles and some reasoning.

When you ask a question, the model writes its answer one token at a time, choosing likely continuations. That’s why answers sound fluent but can still be wrong: it predicts plausible text rather than looking facts up, which can lead to hallucinations.

“Large” refers to the number of parameters, often billions. Smaller, cheaper versions are called small language models.