# Context window

> For the complete documentation index, see [llms.txt](https://attensira.com/llms.txt).
> Every page on this site is also served as Markdown: append `.md` to any URL, or send `Accept: text/markdown`.

> A context window is how much text a language model can read at once, in tokens. Current sizes, why bigger windows still select, and what that means for pages.

- Canonical page: https://attensira.com/glossary/context-window
- Family: glossary
- Last updated: 2026-10-04
- Author: Karl-Gustav Kallasmaa (https://attensira.com/about)
- Claims as data: https://attensira.com/claims.json?slug=context-window

**In short:** A context window is the maximum amount of text, measured in tokens, that a language model can take in at once: the instructions, the conversation so far, any retrieved documents and its own answer. Current large models accept around a million tokens, but answer engines still select a few passages to read, so your page competes section by section.

| Figure | What it counts | n |
| --- | --- | --- |
| 33.5 | sources per ChatGPT answer | 78,667 citations, n=2,349 answers |
| 10.0 | sources per Perplexity answer | 21,713 citations, n=2,175 answers |
| 78.4% | of all citations from ChatGPT | 78,667 of 100,380 citations |

Window 2026-09-05 to 2026-10-04. attensira.com workspace, read over the Attensira MCP on 2026-10-04.

_Source: attensira.com workspace, read over the Attensira MCP, as of 2026-10-04 (n=4,524 answers)._

## Key facts

- Unit: [Tokens](/glossary/tokens)
- Holds: Instructions, conversation, retrieved text, answer
- Current size: About 1M tokens on large Claude and Gemini models
- Not the same as: Memory: the window resets every request
- Related term: [Retrieval-augmented generation](/glossary/rag)
- In the docs: [Measure › Citations and sources](https://docs.attensira.com/measure/citations-and-sources)

## How much material goes into one answer?

More than the answer can use. In the 30 days to 4 October 2026, read over the Attensira MCP, ChatGPT attached 78,667 citations to 2,349 answers to the questions our workspace tracks: about 33.5 per answer. Perplexity attached 21,713 to 2,175 answers, about 10.0 each. A citation here is one URL the engine attached to its answer.

We cannot see inside either engine's context window. From the outside, though, the shape is clear. A ChatGPT answer that carries thirty-odd sources is a few paragraphs long, so most of those pages contribute a sentence at most. Being retrieved into the window is not the same as being used in the answer.

The two engines also differ by a factor of about three on the same questions. A page competing for a ChatGPT answer shares the window with far more material than the same page in a Perplexity answer. These readings cover ChatGPT and Perplexity only; Claude and Gemini answers are not read in this workspace.

## What is a context window?

It is the maximum amount of text a language model can process in one request, measured in [tokens](/glossary/tokens). Everything in the request shares it:

- the system instructions;

- the conversation so far, in [conversational search](/glossary/conversational-search);

- documents retrieved for this answer, in [retrieval-augmented generation](/glossary/rag);

- the answer the model writes.

Anything that does not fit is not seen. The model has no other view of the world in that moment except what it learned in training.

## How large are context windows now?

Large. As read on 4 October 2026, Anthropic's [models overview](https://platform.claude.com/docs/en/about-claude/models/overview) lists:

| Model | Context window |
| --- | --- |
| Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 | 1M tokens |
| Claude Haiku 4.5 | 200K tokens |

Anthropic notes 1M tokens is roughly 555k words on its current tokenizer. Google says many Gemini models accept 1 million or more tokens. These numbers change with every release, so treat any table, including this one, as dated.

## If windows are so big, why do answer engines still select?

Because the web is bigger. A million tokens holds a long book or a mid-sized documentation site. It does not hold the web, or even the top results for one question once you include every page's navigation, scripts and boilerplate. So answer engines still search, [chunk](/glossary/chunking), score and select a few passages per question.

Two more reasons selection persists:

- **Cost and speed.** Every token in the window costs compute and time on every answer.

- **Attention is not uniform.** Anthropic's retrieval research found that passing the top 20 chunks beat passing 5 or 10, but more is not endlessly better: extra material can distract a model. Engines tune how much to include rather than including everything.

Anthropic's own rule of thumb: a knowledge base under about 200,000 tokens can go in the prompt whole. Above that, retrieve.

## Does every question fill the window the same way?

Not on ChatGPT. Across three questions our workspace tracks, it attached between 21.5 and 28.7 citations per answer, depending on the question. Perplexity stayed close to 10 on all three. How much competing material sits beside your page depends on the engine first and the question second.

**Citations attached per answer, three tracked questions, 5 Sep to 4 Oct 2026**

| Question | ChatGPT | Perplexity |
| --- | --- | --- |
| AI prompt volume tool | 28.7 (1,635 over 57 answers) | 9.6 (520 over 54 answers) |
| AI search visibility API | 28.4 (341 over 12 answers) | 9.8 (118 over 12 answers) |
| LLM SEO tool | 21.5 (258 over 12 answers) | 9.6 (115 over 12 answers) |

_Source: attensira.com workspace, read over the Attensira MCP, as of 2026-10-04 (n=159 answers)._

## What does this mean for your pages?

Your page enters the window in pieces, if at all, and competes with pieces of other pages:

- **One question per section.** A section that answers one thing is a clean candidate.

- **Self-contained passages.** The subject, the claim and its qualifier in the same few sentences, because the paragraph above may not be in the window.

- **Your name in the sentence.** If the selected passage does not name you, the answer can use your fact without crediting you. See [AI brand mentions](/glossary/ai-brand-mentions).

- **Less boilerplate.** Content that renders in the HTML, without walls of navigation before it, is cheaper to read and easier to extract.

## What are the common misunderstandings about context windows?

**"A 1M window means the model reads my whole site."** It reads what the application chooses to send, usually a few passages.

**"The window is memory."** It is per request. Memory features store notes and resend them.

**"Bigger windows make SEO irrelevant."** They make selection less tight, not unnecessary. Retrieval still decides what is read.

**"Fill the page with everything, the model will sort it."** Extra material can dilute the passage that matters.

## How does Attensira use this?

Attensira records which of your pages each engine cites and the sentence of the answer each link supports, so you can see which passages made it from the window into the answer. The crew rewrites the passages that are read but not used, ships each change after your approval, and measures named and linked rates before and after. The goal is to be one of the few sources the answer actually uses, not one of the thirty it attaches. [Citation tracking](/features/citation-tracking) shows how each link is tied to the sentence it supports.

## FAQ

### What is a context window?

It is the most text a language model can process in one request, counted in tokens. Everything shares it: the system instructions, the conversation, any documents retrieved for the answer, and the answer itself. Text beyond the window is simply not seen.

### How large is a context window today?

For current large models, about a million tokens. On 4 October 2026 Anthropic listed 1M tokens for Claude Fable 5.1, Opus 5.5 and Sonnet 5.5 and 200K for Haiku 4.5. Google says many Gemini models accept 1 million or more tokens.

### Is the context window the same as memory?

No. The window is what the model can read in one request. It forgets everything when the request ends unless the application sends it again. Long-term 'memory' features work by storing notes and putting them back into the window later.

### If windows keep growing, is retrieval still needed?

For the web, yes. A million tokens is roughly 555k words on Anthropic's current tokenizer, which is large for a document and tiny for the web. Search engines still retrieve a few passages per question. Anthropic suggests skipping retrieval only for knowledge bases under about 200,000 tokens.

### What does the context window mean for how I write a page?

That your page is read in pieces. An answer engine puts selected passages, not whole sites, into the window. Each section should answer one question on its own, with the subject and qualifiers inside it, so it still works when it is the only part of your page the model sees.