Free AI token calculator for ChatGPT, Claude and Gemini
Paste any text to count its tokens, then see what it costs on 12 current models from OpenAI, Anthropic and Google, per request and per month. Exact OpenAI counts, clearly labelled estimates for the rest, and your text never leaves your browser.
- Exact OpenAI token counts
- 12 models compared
- Prices checked 7 Nov 2026
- Free, nothing uploaded
1Paste your text
Paste a prompt, a document or any text. It is counted in your browser and never uploaded.
2Set your usage
3Your estimate
4Compare every model
Cost of the same text and usage on each model, cheapest first.
| Model | Input tokens | Price per 1M (in / out) | Per request | Per month |
|---|
5See how your text splits into tokens
Each coloured block is one token in OpenAI's o200k_base tokenizer. The first 600 tokens are shown.
How to calculate AI token costs
Three steps, all in your browser.
Paste your text
Paste a prompt, a document or a chat history. Characters, words and OpenAI tokens are counted as you type, using OpenAI’s open-source o200k_base tokenizer.
Set your usage
Choose a model, how long each answer should be and how many requests you send a month. Turn on batch pricing for work that can wait.
Compare and plan
See the cost per request and per month, how much of the context window you use, and all 12 models ranked from cheapest to most expensive.
How accurate are the token counts?
Every AI company uses its own tokenizer, and only OpenAI publishes one that runs in a browser. The calculator labels each count so you know how far to trust it.
| Models | How we count | Accuracy |
|---|---|---|
| GPT-5.5 (OpenAI) | o200k_base, the tokenizer OpenAI’s tiktoken library maps to GPT-4o and the GPT-5 family | Exact for plain text |
| GPT-6 family (OpenAI) | o200k_base. OpenAI has not published which tokenizer the GPT-6 models use | Close estimate |
| Claude (Anthropic) | About 2.5 characters per token, based on Anthropic’s statement that 1M tokens is roughly 2.5M characters on its current tokenizer | Estimate. For an exact count, use Anthropic’s token counting endpoint |
| Gemini (Google) | About 4 characters per token, from Google’s Gemini documentation | Estimate. For an exact count, use the Gemini countTokens method |
Real API requests also include a few extra tokens for message formatting, system instructions and tool definitions, so a bill can be slightly higher than a plain text count.
What changes your AI bill
Token count is only part of the cost. These are the factors that change it most.
Output costs more than input
Output tokens are typically priced five times higher than input tokens or more, so the length of each answer often matters more than the length of your prompt.
Reasoning tokens count as output
Thinking models write hidden reasoning before they answer, and it is billed as output. Add a margin to your output estimate when you use them.
Long prompts can cost more
OpenAI charges 2x input and 1.5x output on GPT-6 and GPT-5.5 prompts over 272K tokens, and Claude Haiku 5.5 costs more over 100K tokens. The calculator applies these rules.
Batch and caching cut costs
Batch processing is half price at all three providers. Prompt caching makes repeated input much cheaper, for example 5% of the normal input price on GPT-6.1 Sol and Claude Sonnet 5.5. The calculator does not include caching.
Introductory prices end
Gemini 4 Argon is on an introductory price until 30 November 2026 and Gemini 3.8 Flash until 31 December 2026. The calculator switches to the standard price automatically on those dates.
Your text stays private
Counting happens in your browser. The tokenizer file is downloaded once from jsDelivr and checked against a fixed checksum before it runs. Your text is never sent to CloudFathom or anyone else.
Prices in the calculator come from the official pages: OpenAI API pricing, Claude API pricing and Gemini API pricing. Always confirm there before you commit to a budget.
Frequently asked questions
What is an AI token calculator?
It counts the tokens in a piece of text and turns that number into an API cost for a chosen model. This one also compares 12 current OpenAI, Anthropic and Google models side by side.
How many tokens is 1,000 words?
In English, roughly 1,300 to 1,800 tokens, depending on the model. Paste your own text to see the exact OpenAI count and estimates for Claude and Gemini.
Is the token count exact?
For GPT-5.5 it is exact for plain text, because it uses the same open-source tokenizer. GPT-6 counts are a close estimate, and Claude and Gemini counts are estimates based on each company’s published ratios. API requests also add a few tokens for message formatting and tools.
Is my text uploaded anywhere?
No. All counting happens in your browser. Only the tokenizer file is downloaded, and it does not receive your text.
Why do Claude and Gemini show different token counts?
Each company uses its own tokenizer, so the same text is split into a different number of tokens. Anthropic’s current tokenizer produces noticeably more tokens than OpenAI’s for the same English text.
Does this include ChatGPT, Claude or Gemini subscriptions?
No. It uses pay-as-you-go API prices. Chat app subscriptions charge a monthly fee with usage limits instead of a price per token.
How often are the prices updated?
Prices were checked on 7 November 2026 on each provider’s official pricing page and are reviewed monthly. The calculator shows a warning if the prices have not been checked for more than 45 days.
Try our other tools
New to AI tokens?
Our plain-English guide explains what tokens are, how text is split into them, why other languages cost more and how to cut your AI bill.