Per-token pricing across every model we serve. No subscriptions, no minimums — only the tokens you actually burn.
