Token & cost estimator
Size up a prompt before you pay for it.
Paste a prompt, an article, or a chat transcript and see the token count plus what one call costs on GPT-4o, GPT-4o mini, Claude Sonnet and Gemini Flash. The estimate is computed in your browser, so the text never leaves the page.
- Runs locally
- Five model families
- Instant results
Try it now
Paste your text and watch the numbers update as you type. Switch models to compare price without re-entering anything.
Token & cost estimator
Paste any text to estimate its token count and the cost of one call across popular models.
- Tokens (est.)
- 0
- Characters
- 0
- Input cost
- $0
- Output cost
- $0
Heuristic estimate: CJK ≈ 1 token per character, other scripts ≈ 1 token per 4 characters. Pricing is indicative — check the provider's official pricing page.
Key Highlights
Input tokens dominate most short requests, while output tokens dominate long generations. Estimating both sides lets you spot the case where a cheap model is not actually cheaper — a long answer on a low-cost model can still outspend a short one on a premium model.
Prices shown here are indicative list prices per million tokens. Providers change them often, and discounts, caching and batch tiers all move the real number, so treat the output as a sizing guide.
Specifications
| Token heuristic | CJK ≈ 1 token per character, other scripts ≈ 1 token per 4 characters |
|---|---|
| Models priced | GPT-4o, GPT-4o mini, Claude Sonnet, Gemini Flash |
| Privacy | Compute only; no network request is made |
| Output cost model | Assumes the reply is as long as the input |
How to Use
- 1
Paste your text
Prompt, article or transcript — anything you are about to send.
- 2
Pick a model
Switch between the priced models to compare the same input side by side.
- 3
Read the two numbers
Input cost is what you will pay to send; output cost assumes a reply of equal length.
| Dimension | Token estimator | Provider dashboards | Spreadsheet |
|---|---|---|---|
| No account needed | ✓ | ✗ | ✓ |
| Works before you send | ✓ | ✗ | ✓ |
| Stays up to date with models | ✓ | ✓ | ✗ |
| Text stays local | ✓ | ✗ | ✓ |
Frequently Asked Questions
Why is the count an estimate?
Browsers have no access to provider tokenizers, so the tool uses the widely used approximation of one token per CJK character and one token per four characters elsewhere.
Which price should I trust?
The provider's official pricing page. This tool is for relative comparison and quick sanity checks.
Does it handle other languages?
Yes. Non-Latin scripts are counted as wide characters, which matches how most tokenizers behave.
Ready to size up a real prompt?
Open the estimator, paste the text, and get both numbers in a second.