Choose the model.
Know the cost.
Available models and their configured input/output rates are shown in the workspace. Connect your own compatible provider or use one offered by the installation operator.
One budget per run
A cloud run reserves the budget you choose. Hosted-model usage and compute are charged from that reservation. Unused credit is returned. Requests that would exceed the remaining limit are refused before being sent.
Your own provider
Your provider bills its model usage directly. Latent Agent still meters cloud compute. External provider estimates count toward your run limit and are shown separately from credit charges.
Applications
Hosting is charged according to the application’s allocated memory and configured hourly rate. Sleeping apps do not accrue runtime charges. The live rate appears before deployment.
Media
Each configured media tool displays a fixed job price before submission. A failed job returns its reserved credit. Model availability and prices depend on the provider configured by the operator.
Open credit & usage