Token Counter & Cost Calculator
This free token counter estimates how many tokens your text uses and what it costs to send to, or get back from, AI models by OpenAI, Anthropic (Claude), Google (Gemini), Mistral and DeepSeek. The token count is an estimate, and the costs use each provider's own price in US dollars per million tokens, checked on 26 September 2026.
Cost Estimate for 0 tokens
* The token count is an estimate worked out from the number of words and characters. Each provider counts tokens with its own tokenizer, so the real number varies with the language and the type of text.
Anthropic says Claude models from 4.7 onwards, including this one, use a newer tokenizer that makes about 30% more tokens for the same text, so this count includes 30% extra.
Prices checked on 26 September 2026 on each provider's own price page: OpenAI, Anthropic (Claude), Google (Gemini), Mistral, DeepSeek. They are the providers' own prices in US dollars per 1M tokens, and the providers can change them.
View all model prices (US dollars per 1M tokens)
What is a token?
AI models read and write text in small pieces called tokens. A token can be a whole word, part of a word or a punctuation mark. Providers charge for the tokens you send (input) and the tokens the model writes back (output), so the token count sets the bill. Anthropic and Google both say one token is about four characters of English text.
How this token counter works
- It counts the words and characters in your text and makes two estimates: words × 1.3, and characters ÷ 4. The token figure is the average of the two, rounded up.
- It does not run each provider's own tokenizer (the program that splits text into tokens), so the count is a guide. The real number varies with the language and the type of text.
- For Claude models from 4.7 onwards (Opus 4.7 and later, Sonnet 5 and the Fable models), it adds 30%. Anthropic says the newer tokenizer in these models produces about 30% more tokens for the same text.
- Costs use each provider's standard price. Discounts for sending work in bulk to run later (batch) or for reusing the same prompt (caching) are left out.
- Models that think before they answer also charge that hidden thinking as output tokens. The tool cannot see it, so the real output cost can be higher.
- Some OpenAI models charge more for the whole request once a prompt is over 272,000 tokens, and Gemini Pro models once it is over 200,000. The tool switches to the higher rate when your text is over that size.
Worked example: a 1,000-word document
Say you paste a 1,000-word report that is 6,000 characters long, spaces included.
- Word estimate: 1,000 × 1.3 = 1,300 tokens
- Character estimate: 6,000 ÷ 4 = 1,500 tokens
- Token estimate: (1,300 + 1,500) ÷ 2 = 1,400 tokens
On GPT-5.6 Terra, at $2 per million input tokens and $12 per million output tokens:
- Sending the report: 1,400 × $2 ÷ 1,000,000 = $0.0028
- Getting 1,400 tokens back: 1,400 × $12 ÷ 1,000,000 = $0.0168
- Round trip: $0.0028 + $0.0168 = $0.0196, about 2 US cents
Sending and getting back 1,000 reports like this a month costs about $19.60. On Claude Sonnet 5 the tool adds 30% first (1,400 × 1.3 = 1,820 tokens), so one round trip at $2 in and $10 out is 1,820 × ($2 + $10) ÷ 1,000,000 = $0.0218.
Questions and answers
How many tokens is 1,000 words?
About 1,330 tokens in everyday English, going by Anthropic's rule of thumb that one token is about 0.75 words (1,000 ÷ 0.75 = 1,333). Claude models from 4.7 onwards use a newer tokenizer: Anthropic says 1 million tokens hold roughly 555,000 words on it, which is about 1,800 tokens per 1,000 words. For a 1,000-word document of 6,000 characters, this tool estimates 1,400 tokens, or 1,820 on those newer Claude models.
Is the token count exact?
No, it is an estimate based on the number of words and characters in your text. Each provider splits text with its own tokenizer, and the real count varies with the language and the type of text. Use the tool to plan and compare costs, and check the usage page in your provider account for the exact figures you are billed for.
Why are the prices in US dollars?
OpenAI, Anthropic, Google, Mistral and DeepSeek publish their API prices in US dollars per million tokens, so the tool shows their own figures without converting them. To see a cost in pounds, use the ToolsForTasks Currency Converter.
What is the difference between input and output tokens?
Input tokens are the text you send to the model, such as your question and any document you paste in. Output tokens are the text the model writes back. Models that think before they answer also charge that hidden thinking as output tokens, which this tool cannot see, so the real output cost can be higher. Output costs more for every model in this tool: GPT-5.6 Terra, for example, charges $2 per million input tokens and $12 per million output tokens.
Which AI models does the token counter include?
The main text models on each provider's price list on 26 September 2026: OpenAI's GPT-6, GPT-5.6, GPT-5.5, GPT-5.4, GPT-5.2, GPT-4.1 and GPT-4o; Anthropic's Claude Fable, Opus, Sonnet and Haiku; Google's Gemini 3 and 2.5; Mistral Large, Medium and Small; and DeepSeek V4. Models the providers have retired, plan to shut down or offer by invitation only are left out, as are some older models still on sale, such as GPT-5.1, and OpenAI's Pro versions.
Related tools and guides
For more on context windows and keeping prompts short, read Understanding Token Limits in AI Models. To check the length of a document, tidy it or split it into smaller parts before you send it, try the Word Counter, Text Cleaner or Text Splitter. To build reusable prompts, use the Prompt Template Generator, and to turn a dollar cost into pounds, use the Currency Converter.