Overview

ModelArts provides two billing modes: yearly/monthly and pay-per-use to meet your requirements in different scenarios. For details about the prices of different billing modes, see ModelArts Pricing Details.

Yearly/Monthly is a prepaid billing mode where resources are billed based on the purchase duration. You get a greater discount if you purchase a longer period. This mode is ideal for scenarios where your compute resource demand is stable during the purchase duration.
Pay-per-use is a postpaid billing mode where ModelArts compute resources are billed based on the usage duration, in seconds. A bill is generated on the hour. This mode allows flexible resource usage based on service requirements, eliminating the need for pre-provisioning and preventing over or under-allocation of resources. This mode applies if resource requirements fluctuate. You only need to pay for what you have used.

Table 1 lists the differences between the two billing modes.

**Table 1** Differences
Billing Mode	Yearly/Monthly	Pay-per-use
Payment Method	Prepaid Billed by the purchase period specified in your order	Postpaid Billed by usage duration
Billing Cycle	Billed by the purchase period specified in the order	Billed by the second. A bill is generated on the hour.
Applicable Billing Item	Compute resources	Compute resources and EVS disks
Applicable Resource Pool	Dedicated resource pools	Public resource pools and dedicated resource pools
Applicable Function	ExeML, notebook, model training, and service deployment	ExeML, notebook, model training, and service deployment
Billing Mode Change	The billing mode can be changed to pay-per-use. The pay-per-use mode takes effect only after a yearly/monthly period expires. Changing the Billing Mode from Yearly/Monthly to Pay-per-Use	The billing mode can be changed to yearly/monthly. Changing the Billing Mode from Pay-per-Use to Yearly/Monthly
Flavor Change	Supported	Supported
Application Scenario	This cost-effective mode is ideal when the duration of resource usage is predictable. This mode is recommended for long-term users.	This mode is suitable when you want more flexibility and control on compute resource usage.