← Back

What is an LLM?

Posted on Wed 17 December 2025 in GenAI

An LLM, or Large Language Model, is a type of artificial intelligence trained to understand and generate human language. It's the technology behind tools like ChatGPT, Claude, and Gemini — the systems that can answer questions, write essays, summarize documents, translate languages, and even write code, all by predicting text one piece at a time.

Breaking Down the Name

  • Large — refers to two things: the size of the model (often billions or even trillions of internal parameters) and the massive amount of text data it's trained on (books, articles, websites, code, and more).
  • Language — the model's domain is human language: text, and by extension, things like code, which follow language-like structure.
  • Model — a mathematical system that has learned patterns from data and can use them to make predictions or generate output.

Put together: a very large, data-trained system built specifically to work with language.

What Does It Actually Do?

At its core, an LLM does one thing: it predicts the next word (or "token") in a sequence of text. Given a prompt like "The capital of France is," it calculates the most probable next word — "Paris" — based on patterns learned from enormous amounts of training text. It repeats this prediction over and over, one token at a time, to build out full sentences, paragraphs, or entire documents.

It's a bit like an extremely sophisticated autocomplete — except instead of suggesting one word, it can hold context across an entire conversation, follow instructions, adopt a tone, and reason through multi-step problems.

How Does It Learn?

LLMs go through a couple of key stages:

  • Pretraining — The model is fed massive amounts of text and learns to predict missing or upcoming words. This is where it absorbs grammar, facts, reasoning patterns, and general world knowledge.
  • Fine-tuning — The model is further trained on curated examples and human feedback so it learns to follow instructions, stay helpful, and behave consistently as an assistant rather than just a raw text predictor.

What Can LLMs Do?

  • Answer questions and explain concepts
  • Write and edit essays, emails, and articles
  • Summarize long documents
  • Translate between languages
  • Write, debug, and explain code
  • Hold multi-turn conversations with context
  • Brainstorm ideas or assist with creative writing

What Are Their Limitations?

  • Hallucinations — LLMs can generate confident but incorrect information, since they're predicting plausible text rather than retrieving verified facts.
  • Knowledge cutoffs — Without external tools, an LLM only knows what was in its training data up to a certain date.
  • No true understanding — LLMs recognize and reproduce statistical patterns in language; they don't "understand" the world the way humans do, even though their output often sounds like they do.

The Bottom Line

An LLM is a large-scale AI system trained to predict and generate human-like text based on patterns learned from vast amounts of data. It's the engine behind most of today's popular AI chatbots and writing tools — powerful, flexible, and useful across countless tasks, but still a prediction engine rather than a truly "thinking" entity. Understanding this helps explain both why LLMs are so capable, and why it's still worth double-checking what they tell you.