SkillPedia Skills Encyclopedia
中文 EN
Today's updates 2026-11-17 UTC

No new commits yet today

Open the update feed

AI & LLMs · Beginner

LLM

Large Language Model

A very large program that learned to predict the next piece of text after reading an ocean of it. Whatever you type, it is guessing what is most likely to come next — writing, translating and answering questions are all the same thing used differently.

AI & LLMs Beginner Appears 1550 times across the site, in 762 entries 312 characters about 1 min read 81 views

TL;DRIn 30 seconds

  1. A program that read an ocean of text and learned what should come next.
  2. Writing, translating and answering look like three jobs; to it they are one — guessing the next piece of text.
  3. It guesses fluently, and it is just as fluent when it's wrong — so check the facts and figures elsewhere.

OverviewIn detail

A large language model is a program fed an ocean of books, web pages and code. What it does is plain enough: look at a string of text and guess what should come next. Give it 'the weather today is really' and it works out that 'nice' has the highest probability. The writing, translating and question-answering you see in an AI are all that same guess stretched out, one step at a time.

The trick works because most human knowledge is carried in language. In learning to continue a sentence it picked up grammar, common sense and even the patterns of writing code along the way. Past a certain size, abilities nobody specifically taught it — writing an essay, reasoning something through — started to emerge.

The well-known models — GPT, Claude, Gemini, DeepSeek — are all large language models, just from different companies with different training and different strengths. Everything you type into an AI tool goes to one of these first, and it generates the reply.

It isn't a database and doesn't keep answers stored ready to hand back, which is why it can calmly invent things that aren't true — what people call hallucination. Nor is it connected to the internet by itself: even if the tool you're using has search bolted on, that fresh material is put there by someone else, not carried inside the model.

One more intuitive mistake: it doesn't truly 'understand' what you're saying, it has just learned very convincingly how words fit together. So the more fluently it talks, the more you should check whether the facts hold up.

MythsCommon misunderstandings

  • Thinking it looks answers up in a reference library It looks up nothing — it continues the text by probability, which is why it can invent sources that don't exist
  • Thinking it says 'I don't know' when it can't answer It has no such switch — it hands you a plausible-looking answer just the same

In practiceWhere you will meet it

  • Turning a tangled sentence into plain language, or the other way round into formal prose
  • Dropping a pile of scattered material in and having it come back as a document you could hand over
  • When a piece of code or a contract clause makes no sense, having it explained line by line

SummaryThe key points

  • It guesses onward word by word from whatever you type
  • Past a certain size it picked up essay-writing and coding on its own
  • At heart it generates text; it does not query a database
  • Which is why it can invent untrue things with a straight face

FAQ

What does LLM mean?

A program that read an ocean of text and learned what should come next.Writing, translating and answering look like three jobs; to it they are one — guessing the next piece of text.It guesses fluently, and it is just as fluent when it's wrong — so check the facts and figures elsewhere.

Where will I run into LLM?

Turning a tangled sentence into plain language, or the other way round into formal prose;Dropping a pile of scattered material in and having it come back as a document you could hand over;When a piece of code or a contract clause makes no sense, having it explained line by line。

What do people most often get wrong about LLM?

Thinking it looks answers up in a reference library — but It looks up nothing — it continues the text by probability, which is why it can invent sources that don't exist;Thinking it says 'I don't know' when it can't answer — but It has no such switch — it hands you a plausible-looking answer just the same。

What are the key points about LLM?

It guesses onward word by word from whatever you type;Past a certain size it picked up essay-writing and coding on its own;At heart it generates text; it does not query a database;Which is why it can invent untrue things with a straight face。

Common pitfalls Fluency isn't accuracy — check the facts and figures against another source.

Also written asYou may also see it written as

large language model LLMs language model

These are all the same concept — the same thing, however it is written in a skill description. Do not mistake different spellings for different ideas.

Entry 2 of 186 · All entries

Keep readingRelated reading

Where to go nextExplore more

Back to the skills themselves

Most of the concepts covered here come from skills that install into your AI client with a single command.