Home / Claude context window: size, limits and how to check yours

Claude context window: size, limits and how to check yours

Claude can hold a lot of text in view, but it is not unlimited, and long chats still degrade. This page covers what Anthropic says about the size and how to keep an eye on yours.

How big is it?

Anthropic's help center article How large is the context window on paid Claude plans? says that, as of 2026-10-08:

• newer Opus and Sonnet models support a 500K-token window on all paid plans when chatting with Claude;
• other models have a 200K-token window;
• in Claude Code, newer Opus models support a 1M-token window, and on some plans you need extra usage enabled to use it.

The window depends on the model you pick more than the plan you pay for. Anthropic updates the models and numbers often, so treat this as a snapshot and check the article. A side-by-side comparison with ChatGPT is on context limits by plan.

Why a long Claude chat still gets worse

A larger window does not make a long conversation free. Every reply is written from the whole chat, so very long chats can become slower and less consistent, and details from far back can be missed. The signs are the same as anywhere: it forgets instructions, repeats rejected ideas and drifts off the task.

Claude does not show a meter

Claude will not show how much of the window you have used. Web token counters estimate because Anthropic does not publish a tokenizer for the browser. ContextMeter counts the messages on the page and labels the Claude figure as an estimate. It also handles the way Claude drops older messages from the page as you scroll, by remembering what it has seen and scanning the whole chat when you let it.

What to do when your Claude chat is getting long

1. Wrap up the current thread.
2. Use the handoff prompt to get a structured summary.
3. Start a new chat with the summary pasted in.
4. For things you need every time, use a Project so they are not repeated in the chat.

Frequently asked questions

How big is Claude's context window?

Anthropic's help center says newer Opus and Sonnet models support a 500K-token window on paid plans in Claude chat, and other models 200K. In Claude Code newer Opus models offer 1M. The window follows the model more than the plan.

Is Claude's window bigger on Max than Pro?

According to third-party guides, no: Pro, Max, Team and Enterprise get the same window for the same model. Check Anthropic's help article for the current rules.

Does Claude warn me before the chat is full?

Claude may tell you when a conversation is too long, but it does not show a running meter while you chat.

How do I count tokens for Claude?

Anthropic does not publish an in-browser tokenizer, so web counters estimate. Anthropic's API has a token-counting endpoint for exact counts.

Last reviewed 2026-10-08. Numbers change often; see the sources linked on this page.