TokenSave
Free AI Token Counter for GPT & Claude
A free token counter that shows tokens, characters & API cost for GPT, Claude and Gemini in real time — in any language.
The same text can cost from 1.03× (Chinese) to 2.44× (Punjabi) the tokens of English — see how we measured 41 languages.
🔍 See how it splits into tokens
Each colour is one token (GPT-4o and later). Boxes with a dashed outline are a single character that takes several tokens.
Colours just split tokens. Same price each: more boxes = higher cost.
Official pricing & tokenizer references:
What is a token?
A token is a chunk of text an AI model reads. In English, 1 token is roughly 4 characters or ¾ of a word.
Why do other languages cost more?
Tokenizers are trained mostly on English, so Korean, Japanese, Chinese and many other languages split into more tokens for the same meaning.
Is my text private?
Yes. This free counter runs locally in your browser. Your text is never sent to any server.
Do non-English prompts use more tokens?
Yes: across 41 languages, the same prompt uses from 1.03× (Simplified Chinese) to 2.44× (Punjabi) the tokens of English on GPT's o200k tokenizer. Sending the prompt in English avoids most of that: the Save tokens button translates it on your device (desktop Chrome / Edge) and asks for the reply in your language.
How to use the token counter
- Paste or type your prompt. The count updates as you type.
- Compare what that text costs on each model, from small, cheap models to the largest ones.
- Press Save tokens to clean up spaces, trim filler and, for non-English text, translate it to English on your device. The counter shows how many tokens and how much money you saved.
How the count is calculated
For OpenAI models the counter runs the real o200k tokenizer in your browser, so the number matches what the API bills. Claude, Gemini and other model families use their own tokenizers, which are not public in the same form, so their counts are estimates calibrated against measured text. Cost is tokens ÷ 1,000,000 × the price per million tokens, using each provider's list price and the date it was checked.
Why the same text costs different amounts
- Language. The same meaning takes 1.03× (Chinese) to 2.44× (Punjabi) the tokens of English. Korean is 1.44×, Japanese 1.79×.
- Input vs output. What the model writes costs 2–6 times more per token than what you send.
- Model. Prices per million tokens range from $0.05 to $10 for input across the models listed here.
Guides
- What is a token? A plain-English guide
- How to estimate your AI API bill before you build
- Why output tokens cost more, and how to use fewer
- Context windows explained
- How to cut the token cost of non-English prompts
- Prompt caching explained
- Write shorter prompts without losing quality
- JSON vs YAML vs CSV: which uses the fewest tokens?
- How GPT's new tokenizer cut costs in 40 languages
- GPT-6 API pricing: Astra vs Sol vs Luna
- LLM API pricing compared: 22 models
- Tokens per word, measured
- How to count tokens for GPT, Claude and Gemini
- AI cost glossary