Context window checker
See how much of a model's context window your text fills.
Runs in this tab. The file stays here.
Using 0 tokens
| Model | Window (tokens) | Used | Left | |
|---|---|---|---|---|
0.00% | 128k | |||
0.00% | 128k | |||
0.00% | 1M | |||
0.00% | 200k | |||
0.00% | 200k | |||
0.00% | 2M | |||
0.00% | 128k |
Token count uses the GPT-4o tokenizer; other families count differently. Window sizes are starting values — edit them to match your provider's current docs.
This runs in your browser. Nothing is uploaded.
About Context window checker
Check how much of several models' context windows your text fills, side by side, and how much room is left. Counted in your browser.
Paste a prompt or a document and the page counts its tokens, then sets that number against a table of models: Window size, a bar showing how full it is, and the tokens left. The bar turns amber past 80 percent and red once you're over, where the Left column changes to read "over by" and the amount. Seven models are listed to begin with, and you can edit any name or window size, add a model of your own, remove rows, or reset the table.
The count uses the GPT-4o tokenizer for everything. Other model families count differently, so the percentage for Claude or Llama is an approximation. You can also skip the text box and type a token count directly, which is handy when you already know the number from an API response or want to ask "would 150,000 tokens fit?".
One thing the page doesn't do is reserve space for the reply. There is no output field. The model's answer shares the same window as your prompt, so if you expect a long answer, type a manual token count that already includes it. The window sizes in the table are starting values that go out of date as vendors change their models; edit them to match the provider's current documentation.
How to use Context window checker
- 1Paste your prompt or document, or type a token count in the box beside it.
- 2Read each model's used percentage and tokens left.
- 3Edit a window size, or add a model, if a figure is out of date.
- 4To leave room for the reply, add its expected length to a manual token count.
What it won't do
- Text is counted with one tokenizer (GPT-4o); other families are approximations.
- Window sizes are editable starting values, not a live feed.
- No output-reserve setting and no image or file tokens.
Common questions
Is the window size up to date?
Not reliably. The table holds starting values from when the page was written, and models change often. Edit the window column to match the provider's docs, or add a row for your model.
How do I leave room for the model's answer?
Type a token count into the manual box that is your prompt plus the length of reply you want. The page then shows what's left against each window. There isn't a separate reserve-for-output setting.
Why does the percentage differ from my provider's number?
Because text is counted with the GPT-4o tokenizer. Another vendor's tokenizer may produce more or fewer tokens for the same text, so use the manual box with the provider's own count when you need to be exact.
Related tools
Why this one doesn't upload your file
There is no server to upload to. This page is a static file, and the work happens in your browser using the same graphics and WebAssembly code that renders every other site you visit. Your file is read from disk into memory, processed, and handed back as a download. Once the page has loaded you can disconnect from the network entirely and it keeps working. The longer explanation