Token counts are estimates (each model tokenises slightly differently). Nothing you paste leaves your browser.
A paragraph ≈ 100, a page ≈ 500, a long answer ≈ 2,000.
Cached input tokens cost ~10% of the normal input price on all three providers.
Cost by model
| Model | Input $/1M | Output $/1M | Per request | Per month | Context |
|---|
About AI Token & Cost Calculator
Paste a prompt, a document or a chunk of code and this tool estimates how many tokens it is and what it would cost to send to the current Claude, GPT and Gemini models — per request and per month, with input and output priced separately and an option for cached prompts. Token counts are estimates: each provider uses its own tokeniser, so the number is calibrated to about four characters per token for English, a little denser for code, and roughly double for non-Latin scripts. Prices are list prices per million tokens for the standard API tier, with the date they were last checked shown under the table. Thinking or reasoning tokens bill as output on every provider, so if you use extended thinking, raise the output estimate. Everything runs in your browser; your text is not sent anywhere.
How to use it
- Paste the text you plan to send to the model and pick whether it is prose, code, or a non-Latin language.
- Set roughly how long the answer will be, how many requests you make per month, and how much of the prompt is a repeated (cacheable) system prompt.
- Compare the cost per request and per month across models; the cheapest for your exact numbers is highlighted.
Frequently asked questions
How accurate is the token estimate?
Within about 10–15% for English prose, which is enough for budgeting. For an exact number, each provider has a token-counting endpoint (Anthropic's is count_tokens) that returns the precise count for a specific model at no charge.
What is a token?
A piece of text the model reads at once: usually a word or part of a word, sometimes punctuation. In English, 1,000 tokens is about 750 words. Code and other languages use more tokens per character because the tokeniser was trained mostly on English.
Why are output tokens so much more expensive than input?
Generating text is much more computationally expensive than reading it, so providers price output at roughly five times the input rate. Long answers, and reasoning or thinking tokens, are what drive a bill.
How does prompt caching change the cost?
If the start of your prompt is identical across requests (a long system prompt, a reference document), providers can cache it and charge about a tenth of the normal input price for the cached part. Set the cached share to see the effect.
Are these prices current?
They are list prices checked on the date shown under the table and refreshed regularly, but providers change pricing and add models often. Always confirm on the provider's pricing page before committing a budget.
Useful alongside this
Free invoice generator · Resume → LaTeX converter
More from Anchit Labs
Free to use, runs in your browser. Made by Anchit Labs · Privacy