Skip to content
Agent YAP

What is it?

Context window

The fixed-size working memory of a language model, measured in tokens. Everything the model can "see" at once (system instructions, the conversation so far, retrieved documents, tool results) must fit inside this window. It is like a desk of a fixed size: you can only spread out so many papers before you must put some away. The context window is the root cause of most memory and context challenges in these articles.

From the AI agent glossary

As the Foundational Modelling glossary puts it

The context window is the maximum amount of text, measured in tokens, that a model can consider at once. It is the model's working memory. If a model has a context window of 8,000 tokens, it can "see" roughly 6,000 words of conversation or document at a time; anything beyond that has to be dropped or summarized.

Larger context windows let a model handle longer documents and remember more of a conversation. Expanding the context window is a major area of research, and it is exactly the problem that the Recursive Language Models paper in folder 02 tackles.


See it in the Foundational Modelling glossary

Where Context window is explained

  1. The Case For Multi-Agent: Anthropic's Research SystemMulti-Agent Systems · The architecture: orchestrator-worker
  2. What Context Engineering Is, and Why It Replaced Prompt EngineeringContext Engineering · Starting from what you already know
  3. The Four Moves: Reduce, Offload, Isolate, RetrieveContext Engineering
  4. How Memory Is Built and RetrievedAgentic Memory · Five families of memory mechanisms
  5. Subagents and Multi-Agent OrchestrationBuilding Coding Agents and Harnesses · Concept explanation

Every term, A to Z