This calculator finds exact duplicate prompt lines and excess whitespace, estimates their tokens as characters divided by four, and multiplies that waste by monthly calls and input price.
Loading calculator...
How to use it
Paste up to 20,000 characters, then enter Calls / month and Input price / 1M. Defaults are 100,000 and $3.
Review duplicates before deletion because repeated wording may be intentional.
The source lowercases, trims, and collapses whitespace when deciding whether nonempty lines duplicate.
Fields explained
Paste your prompt – text inspected locally in the browser.
Calls / month – default 100,000.
Input price / 1M – default $3.
Reading the results
| Result | Meaning |
|---|---|
| Est. tokens | Ceiling of characters divided by four |
| Wasted tokens | Duplicates plus whitespace bloat |
| Duplicate lines | Normalized repeated lines |
| Waste / month | Waste tokens times calls and price |
Four characters count as one estimated token.
Warning: the model tokenizer may differ from this English character estimate.
Semantic repetition is outside the calculation.
The formula
wasteTokens = ceil((duplicateChars + whitespaceWaste) / 4); monthly waste is calls times waste tokens times price divided by one million.
| Detected waste | Source behavior |
|---|---|
| Duplicate line | Counts its characters plus newline |
| Extra spaces or tabs | Collapsed to one space |
Repeated blank lines are reduced only after three newlines.
Info: no prompt text is uploaded by this calculator.
Use a tokenizer for a final price estimate.
Worked examples
Duplicate policy. Remove one repeated 40-character line: about 10 tokens per call disappear.
Whitespace. Collapse 20 excess characters: about 5 tokens disappear.
Default rate. Ten wasted tokens at 100,000 calls and $3/1M cost $3 monthly.
Scale. 100 wasted tokens at 1M calls and $5/1M cost $500. One hundred wasted tokens cost $500 monthly.
Common mistakes
Characters are not exact tokens. Verify with the deployed tokenizer.
Duplicate-looking lines can have different purpose. Read them first.
Semantic redundancy is not detected.
Danger: an empty prompt yields zero tokens and zero waste.
Paste the real deployed prompt.
FAQ
Is the text uploaded?
No. The UI says it runs in the browser.
The source contains no network action.
Are tabs waste?
Extra tabs are collapsed.
They contribute to whitespace waste.
Are case changes duplicates?
Yes. Comparison lowercases lines.
Whitespace is normalized too.
Does it find paraphrases?
No. It finds exact normalized duplicates.
Human review is needed for meaning.
What rate is used?
The entered input price.
Default is $3 per million tokens.
Disclaimer
This is an educational estimate for planning. Results depend on the model, provider, hardware, configuration, and current pricing, all of which change.
The live calculator on the page is the source of truth for its exact fields. Verify against current prices and a small real test before scaling spend.
