Pay for the unit that matches the work.
There is no public rate card. Rates depend on GPU type, share and region, and the console shows them before you commit. What is fixed is the model: per GPU-minute for pods, per token for inference, per hour or per term for TrainPods, per GPU per year for operators.
Four units, one wallet.
Every product bills in the unit that matches how it is used. No idle charges on inference, no minimum commitment on pods, and nothing billed that the console did not show first.
| Product | Unit | What that means |
|---|---|---|
| QuickPods | per GPU-minute | One physical GPU serves several users; you pay for your share, for the minutes it runs. |
| Inference | per token | Input and output tokens, metered per model. Idle endpoints cost nothing. |
| TrainPods | per hour, or reserved | Whole GPUs by the hour; clusters reserved for a term where partner capacity exists. |
| Podstack OS | per GPU, per year | DC Suite, NextGen DC Suite and PodVirt licensed per GPU annually; Token Factory as a one-time license plus maintenance. |
What each product bills.
The unit and the note come from the same source as each product page.
QuickPods
Metered by the minute against your Podstack wallet. Rates depend on GPU type and share; the console shows the estimate before you launch.
About QuickPodsTrainPods
On-demand instances bill hourly from your Podstack wallet. Reserved clusters are quoted for their term; request one and a quote comes back.
About TrainPodsInference
Input and output tokens are metered per model and charged to your Podstack wallet. Idle endpoints are free.
About InferenceDC Suite
DC Suite is licensed per GPU per year. Pilots carry no license fee. Ask for a proposal sized to your fleet.
About DC SuiteNextGen DC Suite
Licensed per GPU per year, on the same terms as DC Suite. Pilots carry no license fee.
About NextGen DC SuitePodVirt
PodVirt ships with DC Suite and is covered by the per-GPU annual license. It can also be licensed on its own; ask for a proposal.
About PodVirtToken Factory
Token Factory is a one-time license with an annual maintenance contract for updates and support. Ask for a proposal.
About Token FactoryGrid
No license fee during a pilot. After conversion, the operator keeps 65 to 80 percent of what pods and clusters routed through Grid earn; licenses are separate and never shared.
About GridStart with credits if you are building something real.
Developers with a real project can apply for starter credits, reviewed by a person. Students and researchers with an institutional email get more.
Licence the platform, or opt into Grid.
Podstack Grid
The optional middle layer. An operator can license Podstack OS and never join, or opt in to receive Podstack demand while setting their own floor price and keeping most of what their capacity earns.
Opt in to receive Podstack demand. You set a floor price and keep 65 to 80 percent of what the capacity earns. Operators who never join still run the full OS.
Questions, answered.
Why is there no public rate card?
Rates differ by GPU type, share and region and move with supply, so Podstack shows the current rate in the console at the moment you launch rather than publishing a table that would be wrong by next week.
How do I see the rate before I commit?
Sign in, pick a GPU and a share, and the console shows the rate and an estimate for the run before you launch; nothing is charged until the pod is running.
What does per GPU-minute mean?
QuickPods meter the minutes a pod runs multiplied by the share of the GPU it uses, so an afternoon of experiments on half a GPU costs an afternoon of half a GPU.
Is inference charged when idle?
No, inference bills per input and output token and an endpoint with no traffic costs nothing.
How is reserved capacity priced?
Reserved TrainPods capacity is quoted per request based on GPU model, quantity and term; longer and larger commitments earn better effective rates.
How do operators pay?
Operators licence Podstack OS per GPU per year, with no licence fee during a pilot, and Token Factory is a one-time licence plus maintenance.
See your rate in the console.
Create an account, pick a GPU and the console shows the rate before you launch. Operators can request a pilot instead.