Glossary
What is a context window?
A context window is the amount of text, measured in tokens, a language model can consider at once: the instructions, conversation, documents and tool results it's working with. Larger windows let agents hold more of a codebase or task. Content outside the window is forgotten unless saved and reloaded.
Tokens
Models read text as tokens: pieces of words. In English, a token is roughly three to four characters on average. Context windows range from tens of thousands to a million or more tokens depending on the model.
Why it matters for agents
Long tasks fill the window with tool outputs and history. When it's full, older content must be dropped or summarised. Agents handle this by:
- Saving notes and progress to files.
- Loading instructions only when needed, as skills do.
- Summarising earlier steps.
- Keeping tool outputs short.
Bigger isn't always better
Very long contexts cost more and can make it harder for the model to focus. Retrieving the right few passages (RAG) often beats loading everything.
Example
An agent submitting to 100 directories keeps a tracker file of results instead of the full page text of every form. Its context stays small, and it can resume where it stopped if interrupted.
FAQ
How big are context windows?
From tens of thousands to over a million tokens, depending on the model.
Do models remember past chats?
Not within the model; apps may add memory features that reload notes.
What happens when context is full?
Older content is dropped or summarised, depending on the app.