In one sentence
The context window is the amount of information an AI model can consider at one time when producing a response.
The simple model
When an AI answers, it works from the information in front of it at that moment: the question, any instructions, and any documents or data you provided. That working set is its context. The context window is the limit on how much can fit.
Think of it as the model's desk. It can only read what is on the desk right now. Anything not placed there, it does not know, no matter how relevant.
Why you care
The context window shapes several design decisions.
- It is why RAG exists. Rather than putting an entire library on the desk, you retrieve the few relevant pages and place only those in context.
- It is why documents are split into chunks: so the right small pieces can be selected to fit.
- It affects reliability. Cramming too much in can crowd out what matters, and the model may lean on the wrong part.
More context is not automatically better
It is tempting to put everything in, just in case. But a desk piled with paper is harder to work from than a few relevant pages. Extra, loosely related information can distract the model and make answers worse, not better. Relevance beats volume.
A small example
A team pasted an entire policy manual into every question to "be safe". Answers were often vague or drawn from the wrong section. Switching to retrieving only the relevant few paragraphs made answers sharper and easier to check, and stayed well within the window.
What this does not solve
Fitting information into context does not make it correct or make the model use it well. The model can still misread a passage that is right in front of it. Context is about what the model can see, not whether it reasons correctly.
