Token Counter
Free tools › AI › Token counter
Count tokens and price the call before you spend
Paste a prompt, an article or a whole conversation and see how many tokens it takes and what it costs in each model. Exact count with OpenAI’s own tokenizer, instant estimate for the rest.
- Your text never leaves the tab
- Estimate while typing, exact on demand
- Cost per million tokens
Token counter
Model
0 of 128,000 tokens of context
The exact count uses o200k_base (OpenAI). Claude and Gemini have their own tokenizer, so the figure can differ by about 5%. Prices are reference values: check them on the pricing matrix.
What exactly are you measuring
Models do not read letters: they read chunks. A token is one chunk of a word, and every bill is built on them.
The text is split
The tokenizer cuts by the most frequent pieces of the language. “information” may be 3 tokens; “the” is 1.
Input and output are counted
Almost every provider charges input and output differently: output usually costs 3-5 times more.
It is measured against the context
If the total passes the model window, the conversation is cut or summarised. Here you see the margin left.
How many tokens is...
Measured with the o200k_base tokenizer on real Spanish and English text. Averages: code and tables cost quite a bit more.
| Text | Spanish | English |
|---|---|---|
| 1 word | about 1.25 tokens | about 1.15 tokens |
| 1 page (500 words) | about 620 | about 570 |
| 1,000-word article | about 1,230 | about 1,130 |
| 20 pages (10,000 words) | about 12,300 | about 11,300 |
| A tweet (280 characters) | about 60 | about 55 |
| A 50-page PDF | about 15,400 | about 14,100 |
Frequently asked questions
- How many tokens is 1,000 words?
- Around 1,230 in Spanish and 1,130 in English, measured with o200k_base. It depends on word length and punctuation: 1,000 words of code or of a table cost noticeably more than 1,000 words of prose.
- Is the count exact?
- For OpenAI, yes: it uses the same tokenizer as the model (o200k_base), downloaded only when you press the button. Claude and Gemini do not publish their tokenizer, so there you get an estimate, usually within about 5%.
- Is the text I paste stored anywhere?
- No. There is no server: the text is processed in the tab and disappears when you close it. The only download is the tokenizer, which stays in the browser cache.
- Why is output more expensive than input?
- Because generating token by token costs more than reading them. On most models output costs 3 to 5 times more per token, which is why the conversation column weighs so much.
- Can I use it to budget an API integration?
- Yes, to get an idea before launching a batch. The prices shown are reference values; the pricing matrix keeps the verified ones with their date.
More free tools
The rest of the lab, no sign-up and also inside your browser.