Token WiseAdd to Chrome
Home

How Token Wise calculates cost

Every estimate uses a transparent formula based on prompt length and a model's published token rates. The estimate is calculated locally in your browser.

The estimate, step by step

  1. Estimate input tokens from character count using the extension's token estimation heuristic.
  2. Apply the selected model's input rate. The extension also estimates output cost based on the expected response size.
  3. After optimization, compare the original and rewritten prompt estimates. The displayed savings are the difference between those estimates.
estimated cost = estimated tokens ÷ 1,000 × published rate

What the estimate means

Token Wise's estimates are a practical guide, not a provider invoice. Tokenizers differ between models, code and non-English text can vary, and output length is not known until a model responds. If you use a subscription plan, the estimate is an equivalent API-cost proxy, not necessarily a charge on your plan.

Your prompts stay under your control

Cost estimation happens on your device. Your prompt is sent over HTTPS only when you choose Optimize. A local rule-based rewrite is available as an offline fallback. Optional API monitoring is off by default and observes metadata only.

Token Wise maintains model rates as providers change their pricing. For the latest details, see the privacy notice or contact support@tokenwise.me.