How usage is metered
Gradient meters two kinds of usage, and the rule behind them is to bill the constraint, not the resource.- Voice: per minute. A phone call has a person on the line, so time is the natural limit and the meter. You are billed for connected call time, on both inbound and outbound calls.
- Chat, API, and evaluation: per token. Off the phone there is no clock, so usage is metered by the tokens your agents consume. This covers dashboard and API turns, redteam runs, and rubric grading.
Voice rate by model class
The per-minute voice rate is set by the model class of the published version, stamped at the moment you publish:- Standard: lighter, faster models, billed at the lower per-minute rate.
- Pro: higher-capability models, billed at the higher rate.
The exact current per-minute and per-token rates are shown on your Billing page in the dashboard and on the Gradient pricing page.
Trials, credit, and payment
New organizations are card-first: add a payment method (or redeem a promo code that grants trial credit) before you can run agents. Add a card through hosted Checkout from the Billing page, and manage your card, subscription, and invoices through the hosted customer portal. Promo codes add trial credit and are redeemed on the Billing page.Billing states
An organization is always in one of three states:
When a trialing organization with no card reaches its credit, it is suspended, and requests to run its agents return
402 until a card is added. Connecting a card lifts the suspension immediately.
The billing page
From Billing in the dashboard you can see:- the current billing state and whether a card is connected,
- voice minutes used, split by Standard and Pro class,
- remaining trial credit and any promo codes you have redeemed,
- your outstanding balance, the upcoming invoice preview, and past invoices with hosted links and PDFs.
Organizations and API keys
The tenant that owns billing, members, and keys.
Voice
How phone calls run, the usage voice billing meters.