Cost accounting

Sakana Fugu orchestration tokens are extra work, not a footnote

OpenAI-shaped field names make it easy to treat token_details as a breakdown of the totals you already counted. Official Sakana pricing documentation says the opposite for Ultra: orchestration fields are additional real usage and are billed at the matching input, cached, or output rate.

Checked August 17, 2026Worked from the official usage example

1. Store the official object, not a guessed subset

The live pricing and models pages publish this shape for Ultra (and, on the Models page, for Cyber as well):

usage
{
  "input_tokens": 120,
  "output_tokens": 80,
  "total_tokens": 200,
  "input_tokens_details": {
    "cached_tokens": 0,
    "orchestration_input_tokens": 0,
    "orchestration_input_cached_tokens": 0
  },
  "output_tokens_details": {
    "orchestration_output_tokens": 0
  }
}

Official field meanings:

  • input_tokens: tokens from the user input sent to the first model
  • cached_tokens: cached input for that user input
  • orchestration_input_tokens: sum of input tokens used for orchestration
  • orchestration_input_cached_tokens: cached orchestration input
  • output_tokens: tokens in the final output
  • orchestration_output_tokens: output tokens from orchestration
  • total_tokens: total including orchestration

2. The billable totals are sums, then rates

Billable input = input_tokens + orchestration_input_tokens

Billable cached input = cached_tokens + orchestration_input_cached_tokens

Billable output = output_tokens + orchestration_output_tokens

Estimate = input/1M × input rate + cached/1M × cached rate + output/1M × output rate

Ultra’s published standard rates on August 17, 2026 are $5 input, $0.50 cached, and $30 output per million tokens. Above 272K context the published rates become $10 / $1.00 / $45. Apply the higher tier only when that request’s context crossed the threshold.

3. A worked Ultra example with the same rates the calculator uses

Suppose one standard-context Ultra request reports 1,000,000 user input, 250,000 cached input, 250,000 final output, 2,000,000 orchestration input, 500,000 orchestration cached input, and 500,000 orchestration output.

  • Billable input = 3,000,000 → $15.00
  • Billable cached input = 750,000 → $0.375
  • Billable output = 750,000 → $22.50
  • Total = $37.875

If you had ignored orchestration, you would have stored $12.625 and argued with finance later. The Ultra calculator is that arithmetic with the fields labeled. It still cannot price standard Fugu or Cyber.

4. Operational rules that prevent a second argument

  • Persist the raw usage object beside the computed currency amount and the rate version you applied.
  • Do not assume total_tokens is priced at one blended rate. Input, cached, and output have different prices.
  • Do not treat Ultra max_output_tokens as a spend cap. Official docs say it limits the final model, not the orchestrator.
  • Split mixed traffic: requests under 272K and requests over 272K are two calculations.
  • If a partner route does not return these fields, you cannot reuse this formula without new evidence.

Keep the raw JSON even when the computed dollar amount looks tidy. A later pricing-page change, or a later realization that cached orchestration was omitted, is cheaper to fix from stored fields than from a screenshot of a total.

Sources