Pricing Overview
The platform uses a prepaid credit model: top up first, and charges for each resource are deducted from your credits in real time. You can check your current balance at any time in the top-right corner of the page.
How each resource is billed
Public model endpoints (per token) Billed by actual input/output token usage; billing varies by model type, and each model's unit price is listed in the Pricing section of the Models Hub. Public endpoints are also subject to rate limits.
Private model endpoints (by resource hours) Billed by usage hours for the GPU/CPU resource specification chosen at deployment. The deployment panel shows the estimated cost; enabling Scale-to-Zero under auto scaling can reduce cost during idle periods.
Pods (by resource hours) Billed by usage hours for the selected resource specification. When selecting resources, the cards and the deployment summary show the estimated hourly/monthly rate. Billing stops once the Pod is deleted.
Volume storage (capacity × duration) Billing starts as soon as a volume is enabled, calculated as capacity × usage duration; the unit price varies by storage cluster. After creation, a volume can only be expanded, never shrunk. Billing stops after deletion.
Trying models in the Playground is also billed.
Cost-control tools
- API key budget limits: set a spending cap per key; requests are rejected once the cap is reached. See API keys
- Pre-deployment estimates: both Pods and private endpoints show estimated hourly/monthly costs during the deployment flow
- Real-time visibility: each resource's actual spend is available in its details panel and in the usage reports
- Credit history: every transaction is recorded in the Overview tab of Billing & Payments