- The tool checks whether your document fits in an AI model’s context window.
- You can paste text, upload a file, or estimate from a page count.
- Each model receives a status : Fits, Barely Fits, or Does Not Fit.
- Your text stays in your browser and is never sent anywhere.
An AI’s context window is the amount of text a model can read at once. This free calculator tells you if your document fits in Claude, GPT, Gemini, or another model: paste text, drag a file, or specify a page count, and see at a glance which models can take it in full, with the remaining margin for the response.
How to Use It
- Paste your text, drag a text, Markdown, or code file, or choose ‘Estimate without text’ to specify a number of pages, words, or lines of code. Your text stays in your browser: it is never sent
- Set the tokens to reserve for the model’s response (4,000 by default), then filter by provider, maximum price, or only models that accept images
- Read the verdict and the table: each model is labeled ‘Fits’, ‘Barely Fits’ (beyond 80% of context), or ‘Does Not Fit’. Click a header to sort, or click ‘Copy link with my settings’ to share your simulation: the link never contains your text
What is an AI Context Window?
The context window, or maximum context, is the token limit a language model can process at once: your question, the documents you give it, the conversation history, and its own response all fit within it. A token corresponds to a piece of a word; in French, a word averages about 1.5 tokens. If the total exceeds the window, the model cannot read everything.
How Many Pages Fit in 128,000 Tokens?
Using the tool’s default assumptions (500 words per page, about 1.5 tokens per word in French), 128,000 tokens represent about 165 pages, and one million tokens about 1,300 pages. In English, a word costs fewer tokens: about 190 pages fit in 128,000 tokens. These are rough estimates: code, tables, and technical texts consume more tokens than prose. To measure a text precisely, use our token counter.
Advertised Size vs. Usable Size
The context size advertised by a provider is a technical maximum, not a quality guarantee: on very long contexts, answer quality can degrade. For a document near the limit, prefer a model that keeps a comfortable margin, which is what the ‘Barely Fits’ status indicates.
What to Do If Your Document Doesn’t Fit?
- Split the document into parts processed separately, then combine the responses
- Cascade summarization: summarize each part, then summarize the set of summaries
- Search within documents (RAG): send the model only the passages relevant to the question, instead of the entire document
Before choosing, also compare the cost: the API cost calculator estimates a query, and the model comparator puts models head-to-head.
Limitations to Know
The count is a local estimation accurate to about 20%: each model tokenizes text in its own way, and only the provider’s own counter gives the exact figure. Context sizes, maximum outputs, and prices are public data from OpenRouter, updated daily. Some models apply a different rate beyond a context threshold: the tool indicates this with the note ‘long context tier’.
Frequently Asked Questions
Does my 200-page PDF fit in Claude, GPT, or Gemini?
Choose ‘Estimate without text’, unit ‘Pages’, enter 200: the tool shows the estimated token count and the list of models that can take it in full. At 500 words per page in French, 200 pages represent about 154,000 tokens.
Does the Context Window Include the Model’s Response?
Yes”: “the text sent and the generated response share the same window. That’s why the tool reserves 4”, 000 tokens for the response by default and subtracts them from the margin.
Is My Text Sent Anywhere?
No. The counting happens in your browser: no text is sent to a server or to an AI.