Pricing

Billed per GPU-minute of actual serving. Scale to zero, pay for nothing while idle.

Full pricing is published when you deploy.