AI spend became a real line on the budget. The pricing that drives it deserves the same rigor.

Why Meterary exists

A team building on AI models today is paying real money to a handful of providers whose prices change on their own schedule, with their own naming conventions, and no shared format. Working out what something actually costs means opening several pricing pages, doing the per-million-token arithmetic by hand, and hoping nothing changed since you last checked.

Meterary exists to make that answer available in one place, kept current, for free: the model pricing directory, the calculators, and a workspace for teams that want to see what they're actually spending, not just what a single model costs in theory.

It matters just as much for agents as it does for people. A system deciding which model to call needs machine-readable pricing it can query directly, which is why the whole directory is also a public JSON API.

What we verify, and how

Every model in the directory carries a source link and a last-verified date. Prices are checked against the provider's own published pricing page; where a figure could not be independently confirmed, the model is marked unverified rather than shown as settled fact. As of the last full pass (2026-07-02), 66 of 79 current models are verified against a live source.

What Meterary is not

Meterary is not a billing platform and doesn't touch your provider invoices or API keys — usage is entered manually or imported from a CSV export you provide. It's not a benchmarking service either; prices are the whole focus, not model quality or speed. And a verified price is a snapshot of what we found on a provider's page on the date shown, not a guarantee of what you'll be billed — always check your own invoice against the source link for anything that matters.