A context window is the amount of information a model can work with during one request or conversation state.
The context window is the model’s working space.
It contains the tokens currently available to the model, which may include system instructions, user messages, uploaded content, retrieved information, tool results, and generated text.
Context size is usually measured in tokens.
Working Desk
Think of the context window like a desk. Everything the AI needs for the current task must fit on the desk. A larger desk can hold more documents at the same time.
If the desk is full, some information must be removed, summarized, or retrieved again.
A user has a very long chat.
As more messages are added, more context is used.
Input: A long PDF or document.
System: The text must fit into the available context or be split/retrieved selectively.
Output: Answer based on the visible context.
Instructions + repository files + tool results + chat history can all consume context.
Context Window: The information available for the current model operation.
Memory: Information stored or retrieved across interactions.
Context Window: Information supplied now.
Model Knowledge: Patterns learned during training.
Context Window: Covers available input plus generated tokens according to the model/API design.
Output limit: May be separately constrained.
Explore next
The context window is the amount of information a model can work with at one time.
A short visual explanation of context windows.