AI glossary for businesses
Context window
In short
A context window is the maximum amount of text an AI language model can take into account at one time, measured in tokens. It includes your instructions, any documents you add, the earlier conversation and the model's own reply. Anything beyond that limit is cut off or ignored, so the model effectively cannot see it.
How a context window works
Picture a desk. Everything the AI works with has to fit on its surface: your question, the pasted contract, the last twenty messages and the space for its answer. When the desk is full, the oldest papers slide off the edge. The model does not "remember" them any more, even if they are still visible on your screen.
The size of the desk is measured in tokens, small chunks of text. Context windows differ a lot between models and plans, from a few thousand tokens to hundreds of thousands, and some models now accept around a million. Check the vendor's documentation for the exact figure of the model you use, because it changes often.
A context window is not long-term memory. When you start a new chat, the desk is cleared, unless the tool has a separate memory or project feature.
In practice for a small business
A small wholesaler wants to compare a new 80-page supplier agreement with last year's version. With a model that has a large context window, they can upload both files in Claude or ChatGPT and ask for a list of changed clauses. With a smaller window, the tool may quietly read only part of each file, and the comparison looks complete when it is not.
A second common case: a long chat in which you refined a newsletter for an hour. After a while, the AI starts ignoring the tone rules you gave at the start. That is often the window filling up. A simple fix is to ask for a short summary of the agreed rules and paste it into a fresh chat.
Why it matters and what to watch out for
Attention is uneven: even within the window, models can miss details buried in the middle of very long inputs. Ask about specific sections and verify.
Cost: through an API, you pay for every token in the window on every request. Long documents sent again and again add up.
Bigger is not always better: for large document collections, RAG often works better than stuffing everything into one prompt.
Data: a larger window tempts people to upload whole files. Remove personal data you do not need, in line with GDPR.