No processing of your file contents takes place on any server. You need no data-processing agreement with gottrix to handle confidential or business documents.
Hosted in GermanyGlobal Content Delivery
Origin servers at Hetzner in Germany, delivered worldwide through a content delivery network. Your file contents are never transmitted in the process - they never leave your device.
Language models do not count words but tokens: small pieces of text that differ widely by model, language and script. An English word is often one token, a long German compound several, a Chinese character usually its own. This calculator separates the parts of a chat - system instruction, user message, previous history and text from attachments - and adds image, audio or tool tokens separately, so you can see which part fills the context window.
There are two ways to get the token count. The most accurate is to enter the number your provider's tokenizer or billing reports. Alternatively the tool estimates from the text: it splits it by writing system (Latin, Cyrillic, Arabic, Indic scripts, Han, Kana, Hangul, emoji, digits, punctuation) and shows a range next to the estimate. The upper bound is never larger than the UTF-8 length of the text. The result is explicitly marked as an estimate, because without a model's real tokenizer there is no exact value.
You calculate costs with your own tariff: prices per one million tokens for input, cached input and output, the billing currency, a tax rate and the number of requests. A price of 0 counts as free, an empty field as a missing price - the total then stays visibly incomplete instead of silently showing zero. Currency, tax assumption, tariff region and as-of date appear in the result. There is no conversion and there are no built-in provider prices; your text stays in the browser and is neither uploaded nor stored.
Free, no account - calculate as often as you like; your inputs stay on your device.
Specifications
Specifications
Input formats
Form inputs (no file)
Processing
Locally in your browser (JavaScript)
File upload
None
In 3 steps
Choose the token source: enter your own token count or paste the system instruction, message, history and attachment text.
Enter the output reserve, context window, number of requests and your tariff with currency and tax rate.
Read the context usage, token range and costs before and after tax.
Limitations: Without a model's real tokenizer, the token count from text is an estimate with a range; it can differ markedly for Chinese, Japanese, Korean, Arabic, Indic scripts and emoji. Image, audio and tool tokens are only as accurate as the number you enter from your provider. No public price lists are built in and no currency is converted. The result is not a binding invoice.
FAQ
Is my prompt uploaded?
No. Counting and calculating run entirely in the browser; prompt, history and attachment text never leave your device and are not stored.
Why is a token not the same as a word?
Tokenizers split text into frequent character sequences. A short English word is often one token; long words, rare names, numbers or scripts without spaces produce several. That is why the tool estimates by writing system and never from the word count.
How accurate is the estimate?
It is an approximation with a visible range, not the value of a specific model. For a reliable figure, enter the token count your provider reports; the tool then calculates exactly with your number.
What does a price of 0 mean, and an empty price field?
A price of 0 means free and gives zero cost. An empty field means the price is unknown: if tokens fall under it, the result shows "price missing" and the total stays incomplete.
When does a request fit the context window?
When input tokens plus the output reserve are at most the size of the context window. Exactly at the limit it still fits, one token more does not; from 80 percent the calculator shows "tight".
AI Token Calculator: free and without an account?
Yes. Using it is free and needs no account: you can calculate as often as you like, and your inputs stay on your device.