Serverless Pricing

​

Pricing Items

• Computing Resources: On-Demand. For Serverless, you pay for the actual computing resources consumed; no charge is incurred for unused resources.
• Container Disk: On-Demand. A certain amount of free capacity is supported. Charges will apply to any usage beyond the free allowance, based on capacity and duration.
• Volume Disk: On-Demand. A certain amount of free capacity is supported. Charges will apply to any usage beyond the free allowance, based on capacity and duration.
• Network Volume: On-Demand. Charged separately based on capacity and duration of use.

​

Pricing Method

Charges are calculated based on the GPU specifications configured when creating the Serverless Endpoint, according to the runtime duration of the Workers.

The billing formula is as follows:

Serverless Endpoint Cost = Total cost generated by all Workers running on that Endpoint;

Worker Cost = Worker Running Duration (time in running state, in seconds) * Worker Unit Price (in $/second).

​

Details

• The Worker unit price is determined by the GPU specifications assigned to that Worker. And you can check the pricing on the Serverless GPUs Console.

• If a Worker is configured with multiple GPUs, the unit price of that Worker will change; the final unit price on the configuration page shall prevail.

• Also, after creating a Serverless Endpoint, the actual pricing can be checked on the Serverless Endpoint management page.

​

Account Delinquency Policy

​

Handling of Delinquent Accounts

After your account becomes delinquent (insufficient credits in both account balance and vouchers), the platform will send you notifications and suspend the services. The specific impacts are as follows:

• Creation of new Serverless Endpoints is not supported;
• Only “viewing” or “deleting” of existing Serverless Endpoints is supported, modifications are not allowed, wherein:
  • Workers in running state will no longer accept new requests, but Workers currently processing requests will continue to run until all existing requests are completed;
  • Running Workers will be automatically released after processing existing requests;
  • Eventually, the number of Workers in the Serverless Endpoint will scale down to 0, and no new Workers will be created.

​

Service Restoration

When your account returns to a non-delinquent status, existing Serverless Endpoints will automatically scale up according to their configurations and resume service.