Tool
Document Fit Checker
A document that fits in the window is only half the question — the reply has to fit too. Paste your document, say how long the reply should be, and see per model: does it fit, by how much, what one call costs, and how many chunks it splits into if it doesn't.
Does it fit?
Paste text to check.
| Model | Total needed | Window used | Fits? | Cost / call |
|---|
Token counts use this site's estimator (1.33 per English word, 0.6 per CJK character, 0.25 per other character) — the same numbers as the Token Counter. The fit rule here is input + expected output inside the window, which is an approximation: some models cap input separately, and Gemini shares one pool for input and output. For an exact answer, check the provider's own limits before a production call.
What "fits" means here
Two things have to fit, not one. The input side is your document plus everything else in the prompt — system prompt, instructions, examples. The output side is the reply you expect. A 900,000-token document in a 1,000,000-token window leaves 100,000 tokens for the reply: enough for a summary, not for a full rewrite. The table shows the total (input + expected output) against each window, so the "fits" column already accounts for both.
When it doesn't fit
Three moves, in order. First, shrink the input: cut boilerplate, drop appendices, or summarize the document in stages. Second, split it: the chunk line above tells you how many 300-token pieces the document becomes — the same chunk size the RAG Text Chunker uses, so you can go straight from "doesn't fit" to a split that does. Third, pick a bigger window only if the first two fail: a bigger window is the most expensive way to fit a document, and the cost column shows exactly how much more it runs per call.
The reply cap is separate
Even when the window has room, a single reply has its own ceiling: 128,000 tokens on GPT and Claude, 65,536 by default on Gemini. If your expected output is above that cap, the row says so — the fix is to ask for a shorter reply or split the output across calls, not to find a bigger window.
Next: paste the same document into the Token Counter for the exact token breakdown, or split it with the RAG Text Chunker.