A large language model (LLM) is a transformer trained on enormous text corpora to predict the next token. This simple task makes it acquire general abilities: writing, summarizing, translating, reasoning. It can produce plausible but false answers (hallucinations) and only knows what it learned or what you provide in the context.