Context Window Calculator

How many documents, PDF pages or lines of code fit in Claude, GPT and Gemini?

Model

Custom window, K tokens

Reserve for reply and system prompt, tokens

I have

Language of the text

Optional: paste your own text to measure it

What fits in this model

Every model at a glance

Model	Context	Usable	Your input	Used	Fits?	PDF pages	Lines of code

Compare the cost of these models →

Context windows as of September 2026.

All figures are estimates: a token is roughly 0.75 English words, one A4 page of text is about 670 tokens, a PDF page about 700, a scanned or image-heavy PDF page about 1,600, a line of code about 12 and an image about 1,300 (vendors differ). Non-English text needs more tokens per word, and Claude models from 4.7 onward use a tokenizer that produces about 30% more tokens for the same text, which is reflected here. Always leave room for the reply: models often perform best well below their maximum window. Everything is calculated in your browser; your text is never sent anywhere.

About context windows

A context window is the amount of text a language model can consider at one time: your instructions, the documents you attach, the conversation so far and the reply it is writing all share the same budget, measured in tokens. Windows have grown from a few thousand tokens to a million or more, which is enough for several books or a whole mid-sized codebase.

Knowing whether your material fits saves time and money. A long contract, a research paper set or a repository either goes into a single prompt or has to be split, summarised or searched in pieces. Text pages, scanned PDF pages, source code and images all consume tokens at different rates, and languages other than English generally need noticeably more tokens for the same meaning.

A bigger window is not automatically better: very long prompts cost more, respond more slowly, and models can miss details buried in the middle. Use this calculator to see how much headroom you have on Claude, GPT, Gemini and other models, then decide whether to send everything at once or feed the model only the parts that matter.