Tech Terms Explained Open in the app →

Large Language Model

AI & Machine Learning · Beginner · 5 min read

What is it?

A large language model is an AI trained on huge amounts of text that can understand and generate human-like language by predicting likely next words.

Explain like I'm 5

An LLM is like an extremely well-read autocomplete: it has seen so much writing that it can continue almost any sentence in a sensible, helpful way.

Why was it created?

Earlier language tools were narrow and brittle. LLMs were developed to handle open-ended language tasks — answering, summarizing, translating, coding — with one general model.

Where is it used?

  • Chat assistants
  • Summarization and translation
  • Coding help
  • Drafting and rewriting text

Why should developers care?

LLMs power chat assistants and AI features showing up across products, so most developers now build with or around them.

How does it work?

An LLM is trained to predict the next token (a chunk of text) over enormous datasets. By repeatedly predicting tokens, it generates fluent responses to prompts, guided by patterns it learned.

Real-world example

You ask an assistant to summarize a long article; the LLM reads your prompt and generates a concise summary one token at a time.

Common use cases

  • Conversational assistants
  • Summarizing and extracting information
  • Writing and editing help
  • Code generation

Advantages

  • Handles open-ended language tasks
  • One model, many uses
  • Fluent, flexible output
  • Fast to prototype with

Disadvantages

  • Can confidently state false things (hallucinate)
  • Limited by its context window
  • Knowledge can be out of date
  • Costly to run at scale

When should you use it?

For open-ended language tasks like answering, summarizing, drafting, or coding help.

When should you avoid it?

When you need guaranteed-correct facts or exact calculations without verification.

Alternatives

Rule-based text systemsSmaller task-specific modelsSearch engines for factual lookup

Related terms

TransformerTokenPrompt EngineeringRetrieval-Augmented GenerationHallucination

Interview questions

Beginner

  • What is an LLM?
  • What does it mean that an LLM predicts the next token?

Intermediate

  • What is a context window?
  • Why do LLMs hallucinate?

Senior

  • How does retrieval-augmented generation reduce hallucinations?
  • How would you evaluate an LLM's output quality?

Common misconceptions

  • "LLMs look things up like a database" — they generate text from learned patterns and can be wrong.
  • "LLMs always tell the truth" — they can produce confident but false statements.

Fun facts

  • LLMs work in units called tokens, which are roughly word fragments.
  • They are built on the transformer architecture introduced in 2017.

Timeline

  • 2017 — The transformer architecture is introduced
  • 2020s — LLM-powered assistants reach mainstream use

Learning resources

Quick summary

An LLM is an AI trained on vast text that generates human-like language by predicting likely next tokens, powering chat and writing tools.

Cheat sheet

  • Trained on huge text datasets
  • Predicts the next token
  • Great for open-ended language tasks
  • Can hallucinate — verify facts

If you remember only one thing

An LLM generates fluent language by predicting likely next tokens — impressive, but it can be confidently wrong.