Methodology

Price collection

Each record includes the provider's official pricing URL and the date it was last manually verified. Pricing changes are published by editing the local data and rebuilding the static site.

Cost formulas

Token prices are normalized to USD per one million tokens. Cached input is applied only when a model record includes a cached-input price. Missing prices are represented as unavailable, never silently treated as free.

OpenAI long-context pricing

GPT-5.6 Sol, Terra, and Luna use the displayed Standard short-context prices only through 272,000 input tokens per request. Requests above that threshold use a separate long-context pricing tier. Because V1 does not model that tier, the calculator stops and asks you to consult the official pricing page instead of returning a short-context estimate.

Token estimates

The V1 estimator uses approximately four characters per token. It is explicitly an estimate and not a provider-specific tokenizer.

Corrections

Pricing pages link directly to official sources so users can confirm assumptions and report discrepancies.