Skip to main content
Valkyrie is prepaid and billed per use. Your account has a balance (your quota), and your fine-tuning jobs and deployments draw against it. There are no seats and no charges for idle infrastructure. You manage all of this on the Quota screen; your remaining balance is always shown in the top bar of the dashboard.

​
How the balance works

Your balance is computed from a ledger of every credit (top-up) and debit (usage). Rather than storing a single mutable number, Valkyrie keeps the full transaction history and derives the balance as:
This makes every charge and top-up auditable, and means a correction is recorded as a new entry rather than an edited one.
Because balance is a running total of the ledger, there is no “delete” for a transaction. A reversal is posted as an equal, opposite entry, and both rows stay on the record.

​
Adding funds

Open Quota in the dashboard to view your balance, consumption, and full transaction history. Use Request Quota Increase to ask for more balance.

​
What you pay for

  • Fine-tuning jobs, GPU time for the training run.
  • Deployments, GPU time while a model is running and serving.
Because GPUs are provisioned on demand, you’re charged for the compute you actually use. Stopping an idle deployment (or putting it on a schedule) stops the associated GPU cost.
A running deployment holds a GPU even when it isn’t receiving requests. If you don’t need a model available around the clock, use stop/resume or a schedule to avoid paying for idle time. See Deploy a model.

​
Keeping an eye on spend

  • The Quota screen shows your balance, consumption, and full charge history.
  • The Usage screen shows per-request activity across your deployments.
Administrators can apply internal credit adjustments to a wallet. These are internal balance changes, not payments through a card or invoice.