Context window means the amount of text an AI can keep in view at one time while it answers you: your messages, anything you pasted in, and its own replies. It is measured in tokens, and when a conversation goes past it, older parts get dropped, summarised or refused.
Think of a desk. Everything on the desk, the AI can see and use. The desk has a fixed size. Pile on more papers and something has to come off the far edge, or the AI starts losing track of what is on it.
In a sentence: “I pasted the whole contract, but it was too long for the context window, so the AI only answered about the first half.”
What the word means
The plain point: the context window is the AI’s short-term working space, not everything it learned.
Anthropic, the maker of Claude, describes it in exactly those terms: everything the model can refer to while it writes, including the reply itself, and separate from its training.1
It is counted in tokens, the small chunks of text an AI reads. See token meaning.
A simple example
The plain point: long chats and big files fill the desk.
- You paste in a long report and ask questions. The report, your questions and every answer all sit in the window.
- Hours later, in the same chat, the AI seems to forget a detail from the start. The early part may have been pushed out or summarised to make room.
- Start a fresh chat and paste in only what matters, and it often does better.
Anthropic notes that chat apps such as claude.ai can drop the oldest parts first as a chat grows.1
Why it matters to you
The plain point: a bigger window is not automatically better. What you put in it matters more.
Anthropic’s documentation says accuracy drops as the window fills up, and calls this context rot.1
- Keep it focused. Paste the relevant pages, not the whole archive.
- Start fresh for a new task. Old, unrelated chat history takes up space.
- Sizes change often. Each model has its own limit, and the companies raise them over time. Check the model’s own page rather than trusting an old number. Choosing an AI model helps.
A tidy desk beats a big desk. Give the AI what it needs for this task, and little else.
Often confused with
- Memory: some apps save notes about you between chats. That is a separate feature. The context window only covers the conversation in front of it.
- Training data: the huge amount of text the model learned from before you ever used it. Anthropic’s page draws the same distinction: the window is the model’s “working memory”, not its training.
- Output limit: the longest single reply a model can write in one go. It is a separate limit, and the reply also takes up room in the window.
Questions people ask
What happens when the context window is full?
It depends on the tool. Some refuse or cut off; chat apps may quietly drop the oldest messages to make room, so early instructions can be lost. Anthropic says chat apps such as claude.ai can handle it on a “rolling ‘first in, first out’ basis”, so the oldest text goes first.1
Is a bigger context window better?
It lets you add more, but more is not automatically better: accuracy can slip as the text piles up, a problem called context rot.1
Is the context window the same as memory?
No. Anthropic calls the context window the model’s “working memory”, but it only lasts for the current chat. A chatbot’s memory feature saves notes between chats.2 See AI memory meaning.
Learn more on this site
- Filling the window well: Context engineering and reusable context beats reusable prompts.
- What it costs developers: What you are actually burning.
- Memory and privacy: What happens to what you type.
- The basics: the AI glossary and AI explained for complete beginners.
- Related: How Does ChatGPT Work? What Happens When You Press Send.
- Related: Free Word and Token Counter for ChatGPT, Claude and Gemini.
- Related: Context Window vs AI Memory.
- Related: What Is Attention in AI? Attention Meaning in Plain English.
- Related: What Is Context Engineering? Context Engineering Meaning in Plain English.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What is a context window?
- The amount of text an AI can keep in view at one time
- Everything the AI learned during training
- The size of the chat box on your screen
Show the answer
Answer: The amount of text an AI can keep in view at one time. The context window is the AI’s short-term working space: your messages, anything you pasted and its own replies, counted in tokens.
2. What can happen when a chat grows past the context window?
- The AI keeps the whole chat forever
- Older parts can be dropped or summarised, so early details get lost
- The AI gets faster and more accurate
Show the answer
Answer: Older parts can be dropped or summarised, so early details get lost. When the desk fills up, the oldest parts can be dropped or summarised to make room, so early instructions can be lost.
3. How is the context window different from a chatbot’s memory feature?
- They are two names for the same feature
- The context window saves notes between chats
- The context window lasts only for the current chat; memory saves notes between chats
Show the answer
Answer: The context window lasts only for the current chat; memory saves notes between chats. The context window covers only the conversation in front of the AI; memory is a separate feature that saves notes between chats.
Related terms: Chain of thought · Context engineering · Context rot · Deep research · Few-shot prompting · Grounding
- 1Anthropic, Context windows, undated documentation, read at source 1 Oct 2026“The ‘context window’ refers to all the text a language model can reference when generating a response, including the response itself.” ↑a“Chat interfaces such as claude.ai can also manage the context window on a rolling ‘first in, first out’ basis.” ↑b“As token count grows, accuracy and recall degrade, a phenomenon known as context rot.” ↑c“Chat interfaces such as claude.ai can also manage the context window on a rolling ‘first in, first out’ basis.” ↑d“As token count grows, accuracy and recall degrade, a phenomenon known as context rot.” ↑e
- 2Anthropic, Glossary, undated documentation, read at source 1 Oct 2026the context window “represents a ‘working memory’ for the model.” ↑
The comparisons on this page (the desk, the tidy desk) are this site’s own ways of explaining, not claims from this source.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK