Clarification on request billing calculation

Support Help
1

I’m trying to understand the final cost calculation for API requests.

Here’s an example from my usage logs:

{
  "timestamp": "1768913177593",
  "model": "claude-4.5-opus-high-thinking",
  "kind": "Usage-based",
  "userEmail": "***",
  "requestsCosts": 2,
  "tokenUsage": {
    "inputTokens": 26,
    "outputTokens": 10168,
    "cacheWriteTokens": 76421,
    "cacheReadTokens": 208723,
    "totalCents": 83.632275
  }
}

What amount will be charged to me? $0.83632275 ? How is requestsCosts used in the calculations?
I’ve exceeded the plan usage limit and am currently using On-demand Usage, if that matters.

Thank you!

3

Hey, thanks for the question.

Here’s how it works:

The requestsCosts: 2 field is “request units” from the old billing system, kept for backwards compatibility. But with usage-based pricing, it does NOT affect the cost.

The important field is isTokenBasedCall: true, which means this request is billed using the usage-based model.

When isTokenBasedCall: true:

  • You’re charged totalCents (in your case 83.632275 cents, which is $0.84)
  • requestsCosts is ignored for pricing

How totalCents is calculated in your example:

  • Input: 26 × $3/1M = $0.000078
  • Output: 10168 × $15/1M = $0.152520
  • Cache write: 76421 × $3.75/1M = $0.286579
  • Cache read: 208723 × $0.30/1M = $0.062617
  • Cursor Token Fee: +$0.25/1M on all tokens
  • Total: $0.836… = 83.6 cents

Source: Cursor Docs (requestsCosts is described as “Cost in request units”, and tokenUsage.totalCents is the exact cost for token-based calls)

You’ll be charged $0.84 for this request.