Home / AI context window sizes by plan
AI context window sizes by plan
The context window is how much text a model can consider at once. When a chat gets close to it, older parts are dropped or ignored and answers get worse. This page puts the numbers for ChatGPT and Claude in one place, with the caveats.
Sources disagree and the vendors change these often. The figures below are what public sources reported when reviewed, and the "ContextMeter default" is a conservative starting value you can edit. For anything important, check the vendor's own page.
ChatGPT
| Plan | Reported range | ContextMeter default |
|---|---|---|
| Free | 8K to 27K | 16K |
| Plus (also Go, Business) | 32K to 54K for standard models; 256K for reasoning models | 32K |
| Pro / Enterprise | 128K for standard models; up to 400K for reasoning on Pro | 128K |
Free-plan numbers move the most: older guides say 8K, mid-2026 guides say 16K, and the newest say 27K. The window also depends on which model you use, and reasoning models get larger ones. See the ChatGPT context window guide.
Claude
| Where | Window | ContextMeter default |
|---|---|---|
| Claude chat, newer Opus and Sonnet models, paid plans | 500K according to Anthropic's help center | 200K standard |
| Claude chat, other models | 200K | 200K standard |
| Claude Code, newer Opus models | 1M (some plans need extra usage enabled) | 1M extended |
Claude's window follows the model you pick more than the plan you pay for. Anthropic's article is the one to trust: How large is the context window on paid Claude plans?. See the Claude context window guide.
What the page can't see
The real context also includes system prompts, saved memory, tool and search results, uploaded files and sometimes hidden reasoning. A meter that counts visible messages, including ContextMeter, therefore shows a lower bound. The true figure is higher.
When to act
Quality tends to slip before the hard limit. A reasonable habit is to wrap up or hand off at around 50 to 70 percent. See when to start a new chat.
Frequently asked questions
What is a context window?
The amount of text, in tokens, a model can consider at once. Your prompt, the chat so far and the reply must all fit.
Why do sources give different numbers?
Vendors change windows by model and plan, publish them in different places, and the chat apps can use smaller windows than the API. Third-party guides copy older figures. Treat any single number with care and check the vendor's help page.
Which number does ContextMeter use?
It starts with 16K for ChatGPT Free, 32K for Plus and 128K for Pro, and 200K for Claude, and lets you change any of them or enter your own.
Last reviewed 2026-10-08. Numbers change often; see the sources linked on this page.
ContextMeter is a free, private browser extension that shows how full your ChatGPT or Claude chat is while you use it, and helps you carry the chat into a fresh one. See how it works.