How much will my prompt cost? Count tokens for GPT, Claude and Gemini
Paste any text, this tool shows how many pieces the most popular bots will chop it into: GPT-4o, GPT-4o mini, Claude Sonnet 4.5, Claude Haiku 4.5, Claude Opus 4.7, Gemini 2.5 Pro and Gemini 2.5 Flash. All seven at once, no signup.
A token is a chunk of text the bot treats as a single unit. Usually it's a part of a word, sometimes a whole short word. Rule of thumb: one English word ≈ 1.3 tokens, one non-English word (Spanish, German, Japanese, Polish) ≈ 1.5-2 tokens. The more tokens, the more you pay per question, and the closer you get to the limit on how much the bot reads at once.
Below the numbers you'll see colored blocks, each block is one token. It builds intuition: you'll see how the bot *"looks at"* your sentence, and why non-English costs more than English.
How to use it
- Paste your text in the top box. A prompt for the bot, a snippet of docs, code, a message, anything.
- Below it you get a table: how many tokens your text has in each bot, how many characters per token, and what percentage of the limit you're using.
- The cheapest option (fewest tokens = cheapest call) gets highlighted green, the most expensive red.
- Click any row to see the colored blocks at the bottom, each one is a single token in that bot.
- Click the copy icon in any row to send the token count straight to your clipboard. Handy when you want to multiply it by a price-list rate.
When this is useful
Six typical situations where the token counter gives you a concrete answer instead of a guess:
- Quoting a bot project. A client asks *"how much will the chatbot with this prompt cost me?"*. Paste the prompt, multiply tokens by the provider's price, and you have a real number instead of a guess.
- Checking if your text fits. You have long instructions for the bot, plus chat history, plus documents from a knowledge base. Will it all fit inside Claude's 200,000-token limit? Paste it in and see the answer instantly.
- Comparing languages. Non-English text uses about twice as many tokens as English (the bots learned mostly from English). See exactly how much: "Hola mundo" = 4 tokens, "Hello world" = 2.
- Cost optimization. GPT-4o might use 15% fewer tokens than Claude for the same text. At 10 million questions per month, that's real money over a year.
- **Catching the *"text too long"* error. The bot returned an error saying your input doesn't fit, but you don't know which chunk pushed it over. Paste fragments one at a time and find the piece that broke the limit**.
- Planning a bot that chats with your documents. Say you have 10,000 files averaging 500 tokens each. That's 5 million tokens total. Calculate the cost of splitting them into chunks and converting them to a format the bot understands, know the bill before you spend.