Claude's Context Window, Explained

How much Claude can hold in mind at once, which models give you a million tokens, and why a bigger window isn't automatically better.

Last updated August 2026

Heads up: some links below are referral links. They cost you nothing extra and never change what we recommend. How this site makes money.

The short answer

The context window is how much text Claude can consider at once — your conversation, uploaded documents and its own replies, all together.

Claude Opus 5 and Sonnet 5 support a 1M-token context window on all paid plans. Older models get 500K, and outside those the standard is 200K — roughly 500 pages of text.

ModelContext windowAvailability
Opus 5, Sonnet 51M tokensAll paid plans
Opus 4.8 / 4.7 / 4.6, Sonnet 4.6500K tokensAll paid plans
Other models200K tokensStandard

What a token is, briefly

A token is roughly three-quarters of a word. So:

Everything counts against it: your messages, uploaded files, and Claude's own responses. A long conversation consumes the window even without attachments.

In Claude Code

When using Claude Code on Pro, Max, Team or Enterprise, the 1M-token window is available on Sonnet 5, Fable 5, Opus 5, and the 4.6–4.8 Opus models.

One wrinkle worth knowing: Pro users need usage credits enabled to access the 1M window on Opus models. If you're on Pro and it seems unavailable, that's usually why.

Why bigger isn't automatically better

Two reasons people miss.

It costs you. Every message re-sends the whole conversation. A thread that has filled a large window is expensive per reply, which is a common and invisible cause of hitting usage limits. Starting a fresh conversation when you change topic genuinely saves allowance.

Signal-to-noise. Filling a window with marginally relevant material doesn't reliably improve answers. Ten pages of exactly the right context beats a thousand pages of nearly-right context.

Better than filling the window: put reference material in a Project. Retrieval pulls the relevant parts as needed instead of carrying everything in every message.

On the API

Building on the platform rather than using the app: premium pricing applies to prompts over 200K tokens — $10 per million input tokens and $37.50 per million output.

Worth designing around. Retrieval that sends 50K relevant tokens is dramatically cheaper than sending 500K and hoping.

Common questions

How big is Claude's context window?
1M tokens on Opus 5 and Sonnet 5 for paid plans, 500K on the 4.6–4.8 generation, and 200K as the standard elsewhere.
How many pages is 1M tokens?
Roughly 2,500 pages of text, or a substantial codebase.
Does a longer conversation cost more?
Yes. Each message re-sends the whole thread, so long conversations consume your allowance faster per reply. Starting fresh when you switch topic helps.
Why can't I access the 1M context window on Pro?
On Claude Code, Pro users need usage credits enabled to get the 1M window on Opus models.

Try it on a real document

The free tier is enough to see how Claude handles a long document — that's the fastest way to judge whether the bigger window matters to you.


Related