LLM tokens and cost
LLM token estimator.
Paste text to estimate its tokens from the characters and words in it. This is a rule-of-thumb estimate, not a real tokenizer. Your text stays in your browser.
Tokens are roughly characters / 4 or words / 0.75 for English text. A 12,000-character, 2,000-word draft comes out at about 3,000 tokens by characters and 2,667 by words. At a $2.00 per 1M example input price that is $0.005334 to $0.006000.
Worked example
A 12,000-character, 2,000-word draft. By characters: 12,000 / 4 = 3,000 tokens. By words: 2,000 / 0.75 = 2,666.67, rounded up to 2,667 tokens. At a $2.00 per 1M example price the input costs $0.005334 to $0.006000 (3,000 x 2 / 1,000,000 = $0.006000). The two heuristics disagree by 333 tokens (11.1%), which is the spread you should expect from a heuristic.
How it works
This is an estimate. The 4 characters and 0.75 words per token ratios are a common rule of thumb for English text, not a figure from a provider's pricing page, and you can change both. Anthropic's pricing page says the token count depends on the model's tokenizer, the language and the type of content, and that Claude 4.7 and later models use a newer tokenizer that produces approximately 30 percent more tokens for the same text than earlier models. Real counts come from the model's own tokenizer, and different models can tokenize the same text differently. Anthropic describes its token counting endpoint as an estimate that can differ by a small amount, and OpenAI notes that estimates like characters / 4 are inaccurate for images and files. Use the provider's token counting API when the number has to be right, for example to size a request against a hard limit.
Characters are counted as Unicode code points and words as runs of non-space characters. Code, numbers and non-English text can tokenize differently from English prose (Anthropic says the count varies by language and content type), so change the characters-per-token setting if you have measured a ratio for your content.
Sources and basis (opened 2026-10-08): Pricing - Claude Platform Docs; Token counting - Claude Platform Docs; Counting tokens | OpenAI API. Everything else is arithmetic on your inputs.
All prices on this page are examples you can overwrite, not any provider's current prices. Check your provider's pricing page before you rely on a number. Results are planning estimates, not a quote or financial advice. Everything runs in your browser, and the text you paste is never uploaded or stored.
FAQ
How many tokens is my prompt?
As a rough guide, divide the characters by 4 or the words by 0.75 for English text. A 2,000-word, 12,000-character text is about 2,667 to 3,000 tokens. Only the model's own tokenizer gives the exact figure.
How many words is 1,000 tokens?
Using the rule of thumb of 0.75 words per token, 1,000 tokens is about 750 words. The ratio varies by language and content type.
Is this the same count the API bills me for?
No. It is a heuristic from character and word counts. Billing uses the provider's tokenizer, and request structure such as message roles can add tokens. Use the provider's token counting endpoint for an exact figure.
Is my text sent anywhere?
No. The estimate is calculated in your browser, and the text is not uploaded or stored.