BeginnerLesson 3 of 57 min read
The context window: its working memory
A course from the Academy: They are not all the same.
Go to the quiz ↓It's the feature most often highlighted in announcements, and the most misunderstood.
What it is
The context window is the amount of text it can have in view at once: your question, the entire conversation, and any pasted documents. It's measured in tokens, those chunks of words. One hundred thousand tokens is roughly seventy-five thousand words, or a short novel.
What it changes
A large window lets you paste a long report and ask questions about it, or hold a very long conversation without the start slipping away.
When the limit is reached, the oldest content drops out. It doesn't forget it out of distraction: it simply can't see it anymore. That's why a long conversation eventually "loses" an instruction given at the start.
The nuance nobody mentions
A large window doesn't mean good reading. A model capable of ingesting a million tokens doesn't pay equal attention everywhere: what sits in the middle of a very long text is handled less well than what's at the edges. The stated capacity is a technical maximum, not a guarantee of quality.
What to do with it
Put the essentials at the start or the end of what you paste. And restate the key instruction in your latest message rather than relying on the one from the beginning.
What to try now
In a long conversation, ask it again for an instruction given right at the start. See if it's still there.
Key concepts
Complete the lesson
A 4-question quiz: 3 correct answers are enough, plus an exercise to practise. It opens in the Academy, on this same lesson.
Take the quiz and complete the lesson Free. With an account, your progress is saved and leads to the course certificate.The course: They are not all the same.
- Quick or considered: two families8 min
- Multimodal: when it sees and hears7 min
- The context window: its working memory7 min
- Open or closed: what it means for you7 min
- Choosing in practice, in three questions6 min