Skip to content

Cost estimation

No dollar prices in this chapter. LLM pricing changes on a monthly cadence — anything committed to source will lie in six months. This chapter teaches you:

  1. What the billable units are per service (tokens, images, seconds, requests) — those don't change month to month.
  2. Where the current prices live — the official pricing pages, calculators, and cost-management surfaces per cloud.
  3. How to estimate before you spend — units × current-price arithmetic, plus the vendor calculator that carries the always-fresh number.
  4. How to catch over-runs after you spend — budgets + alerts per cloud.

The Chiron reader owns the prices; the recipe owns the units + the wiring.

Docs verified 2026-08-08

Pages

  • Units — every billable unit per service per cloud (tokens, images, video-seconds, deploy-hours). Stable across price changes.
  • Tooling — official calculators + cost-management + budget-alert wiring per cloud.
  • Estimate — units × your-current-price arithmetic. Includes a small Python helper (examples/cost/service) that you feed the prices to.
  • Deploy — Terraform composing E1 baseline + optional per-cloud budget alerts.

The one rule

If you see a hard-coded dollar in this chapter, that's a bug — file it. Every "example" here uses either the reader's own prices as env vars, or a placeholder like $YOUR_INPUT_TOKEN_PRICE, plus a link to the live pricing page so the reader can fill it in.

Validate-only

No live billing API calls in CI. The examples/cost/service estimator is pure arithmetic on JSON input — it doesn't hit any cloud.