GenAI & LLMs · 1 min read · Updated 2026-10-05
What is a large language model (LLM)?
Guessing the next word
Given 'The sky is', the model sees that 'blue' is very likely. It adds the word, then guesses the next one, and keeps going. That is how it writes whole answers.
Why 'large'?
It has billions of adjustable settings, called parameters, and it learned from an enormous amount of text. More learning usually means more skill.
What it can and cannot do
It can explain, summarise, translate and help with code. It does not look things up by itself, and it can be wrong. Add a calculator or web search and it gets much more reliable.
Key takeaways
- An LLM keeps guessing the next word.
- Its knowledge is stored inside its settings, not in a database.
- Tools like search and calculators fix many weak spots.
Quick questions
What is a token?
A small piece of text, like a short word or part of a word. AI reads and writes in tokens.
What is a knowledge cutoff?
The date its learning stopped. It does not know what happened after that unless you tell it.
Is it really intelligent?
It is very useful, but it works by patterns and makes mistakes. People still debate what to call that.