What this section is
This pricing reference is independent and solo-operated — not affiliated with any provider. It exists because LLM pricing changes constantly (deprecation waves, intro-pricing expirations, silent page edits) and most comparison content goes stale within weeks. Our contract is simple: verified rows only — a price appears here only after being read from the provider's official pricing page, and every row carries its source link and verified date.
The verification pipeline
- Source of truth: provider pricing pages only (OpenAI, Anthropic, Google official docs) — never third-party aggregators. Each model page links its exact source.
- Stamps: every row shows its verified date (2026-08-06 for the current dataset).
- Deprecations: only officially announced shutdowns/price changes are listed, each with the announcement link — no rumors.
- Refresh: rows are re-verified on refresh passes; a stale row that a provider changed under us is the failure mode we optimize against — see below.
The calculator, exactly
Per-request cost = (input tokens × input rate + output tokens × output rate) ÷ 1,000,000; monthly = per-request × requests. Straight rate-card math — caching/batch discounts and tokenizer differences are called out under every result rather than silently modeled, because they vary by workload.
Caught a stale price?
Email hello@answerfootprint.com with the provider page — corrections ship fast and get noted. A stale-price report is the most valuable thing a reader can send us.