# Training

> The process where a model learns patterns from huge amounts of text, building the knowledge it later uses, fixed at a cut-off date.

**Category:** Under the hood

**Training is how a model learns.** It reads enormous amounts of text and adjusts itself until it is good at predicting what comes next. That is where its "knowledge" comes from, not a database it looks things up in.

Two things follow. Training ends on a date, so the model has a **knowledge cut-off** and will not know newer events unless you add them. And what it learned reflects its training data, patterns and all, which is why the source you feed it still matters.

## At work

Ask about a deal that closed last week and a model may not know it: that is the training cut-off. Turn on web search, or paste the details in, so it works from current facts.

## Often confused with

Learning in the moment. Training happened once, before you ever used the model; it does not learn from your chats as you go.

## Related terms

- Model
- Machine learning
- Large language model

## See it in a lesson

[How generative AI works](/en/learn/foundations/how-generative-ai-works)

## Further reading

- [What is machine learning? (IBM)](https://www.ibm.com/think/topics/machine-learning): How learning from data works, training included.
- [Machine Learning Crash Course (Google)](https://developers.google.com/machine-learning/crash-course): A free, hands-on course that covers training.
- [Large language model (Wikipedia)](https://en.wikipedia.org/wiki/Large_language_model): Background on how these models are trained.

---

Last updated: 24/07/2026  
Source: Rollo Academy glossary