AI

What is a large language model (LLM)?

What a large language model is and how it works: tokens, next-token prediction, why answers vary, what LLMs are good at, and why they can hallucinate.

  • 6 min read
  • 5 questions
  • +35 XP
Transcript

LLMs, explained by cats. Kitty opens a chat window. Types: “Write me an email. Explain Python. Summarize this giant wall of text.” And somehow… the machine replies like a very confident cat. That’s an LLM. LLM stands for Large Language Model.

Large, because it trains on huge amounts of text. Language, because it works with words. Model, because it predicts what comes next. That’s the trick. It doesn’t “think” like a human. It looks at your prompt and guesses the next most likely token.

Then the next. Then the next. Token means small chunks of text. Words. Pieces of words. Even punctuation. So if Kitty types, “The cat sat on the…” the model might guess “mat.” Scale that up to billions of patterns, and now it can answer questions, write code, translate text, and sound suspiciously smart.

But. LLMs are not magical. They can be wrong. They can hallucinate. They can sound certain while making total nonsense. So the rule is: use the cat brain too. LLMs are great for drafting, explaining, brainstorming, and coding help. Not great as an all-knowing wizard.