About this tool
Will this document fit in the model's context? It is the first practical question in any LLM project, and answering it means translating between three units: the tokens models actually bill and limit by, the words humans write, and the pages documents are measured in. This calculator converts between all three using the standard English approximation of about four characters or three quarters of a word per token, and five hundred words per page. Presets for current GPT, Claude and Gemini generations let you check any amount against real context windows instantly — see that your 80-page report is about 30,000 words or 40,000 tokens, comfortably inside a 128K window but a third of a 1M window's useful budget with output room reserved. Type a value in any field and the others update live, so it doubles as a quick estimator while writing prompts or chunking documents for RAG. Everything is simple arithmetic done locally; nothing is sent anywhere. Estimates are for English prose — code and other languages tokenize differently.
How to use Context Window Calculator
- Enter a value in any field: tokens, words or pages.
- The other two units convert instantly using standard English ratios.
- Pick a model preset to compare against its context window.
- See what percentage of the window your content occupies.
Frequently asked questions
How accurate is the token estimate?
For English prose, about 0.75 words per token is within 10-15% of real tokenizers. Code, CJK text and numbers tokenize less efficiently.
Why reserve room for the output?
The context window covers input plus generated output. Planning to use 100% of it leaves the model no room to answer.
How many words is one page?
The convention used here is 500 words per page, typical for a formatted document; a dense manuscript page is closer to 250-300.
Do longer contexts cost more?
Yes. API billing scales with input tokens, so filling a 1M window costs roughly eight times filling a 128K one.