Token Counter
Paste a prompt or a document to count its tokens and check it against a context window.
Text is sent only to count tokens and is never stored.
How to use the token counter
Paste your text into the box. While you type, the counter shows a quick estimate based on the usual rule of thumb of about four characters per token. Half a second after you stop typing, it asks our server for the exact number from the tokenizer you picked, and the badge next to the number changes from estimate to exact.
Pick the tokenizer that matches your model:
- o200k_base is used by GPT-4o, GPT-4.1, the GPT-5 family and the o-series reasoning models.
- cl100k_base is used by GPT-4 and GPT-3.5 Turbo, and by OpenAI's text-embedding-3 models.
- Other models shows the four-characters-per-token estimate. Claude, Gemini, Llama, Mistral and other model families use their own tokenizers, so treat the number as a ballpark.
Then choose a context window, or type your own size, to see how much of it your text uses and how many tokens are left for the answer.
What a token is
Language models do not read letters or words. They read tokens: chunks of text that a tokenizer cut from your input using a fixed vocabulary. A common English word is often one token, a rare word can be two or three, and spaces and punctuation are usually attached to the token that follows them. Code, URLs, numbers and languages other than English tend to use more tokens per character, which is why the same length of text can cost very different amounts.
Why count tokens before you send
- Stay inside the limit. Requests that exceed a model's context window fail or get truncated.
- Estimate cost. API providers bill per token, usually priced per million. Enter your model's input price in the optional field to turn the count into dollars. We do not keep a price list, because prices change too often to stay correct.
- Trim long prompts. Comparing counts before and after an edit shows which parts of a prompt are worth keeping.
Limits
The exact count supports text up to 200,000 characters per request and up to 30 requests a minute from one address. Counts for chat requests sent through an API are slightly higher than the count of the text alone, because every message adds a few tokens of formatting. Images, files and tool definitions are counted separately by each provider and are not included here.
Is my text stored?
No. The text is sent to our server only to run the tokenizer and is discarded right after the count. It is not saved, logged or used for anything else.
Why does the number change after I stop typing?
The first number is an estimate calculated in your browser. The exact number comes from the real tokenizer a moment later. For ordinary English prose the two are usually close; for code or other languages the difference can be larger.
Can I count tokens for Claude or Gemini?
You get an estimate. Anthropic and Google use their own tokenizers and offer token-counting endpoints in their APIs. Use those when you need an exact number for billing.
How many words is 1,000 tokens?
For typical English text, roughly 700 to 800 words. The ratio drops for technical text, code and most other languages.
Does the count include the model's answer?
No. It counts only the text you paste. Output tokens are billed separately and usually cost more than input tokens.
AI models tools from our directory
Browse: AI models, Developer tools
Terminal-based browser testing with automated pass-fail proof
All-in-one solution for mobile data management and recovery.
Advanced AI for training and deploying machine learning models.
Automated solution for bypassing online user verification challenges.
Stay informed about the latest AI products and trends.
