Token Counter
Counts with the real GPT tokenizer, right here in the page.
Typed and counted entirely on your device. Nothing is sent anywhere.
Loading the tokenizer…
Counts use the GPT tokenizer (o200k / cl100k). Claude uses a different one, so those numbers are close but not exact. Prices are illustrative and change — check the provider before budgeting.
About tokens
Language models read text as tokens rather than words or letters. Billing, context limits and truncation all work in tokens, so knowing the count tells you what a request will cost and whether it will fit.
What is a token?
A chunk of text, usually part of a word. In English one token averages about four characters, so 100 tokens is roughly 75 words. Common words are a single token; unusual words, names and other languages split into several.
How many tokens is a page of text?
A standard page of about 500 words comes to roughly 650 tokens. A 10-page document is somewhere near 6,500.
Is this accurate for Claude?
Close, but not exact. This uses the GPT tokenizer. Claude uses a different one and typically lands within about 10% for English text — fine for estimating, not for precise billing.
Is my text sent anywhere?
No. The tokenizer is downloaded to your browser and runs there. Nothing you paste is transmitted, which matters if you are working with anything confidential.
Why does the same text cost different amounts?
Models charge separately for input and output, and output is normally three to four times the input rate. The figure here covers input only.