API Gateway:Billing details

Instance fees

Data processing fees

AI Gateway instances support three network types: Internet, private network, or both. Each type is billed separately. Data processing volume per billing cycle equals the total request and response data.

Request CU fees

A Capacity Unit (CU) measures the resource consumption of a Serverless instance.

CU fees are charged hourly. Usage under one hour counts as one hour. Usage is rounded up to the nearest 1,000 CUs, with a minimum of 100,000 CUs per hour.

Hourly CU fee = max{Actual CUs for the hour, 100,000} × CU price

Billing examples:

• If the actual number of CUs used in an hour is 50,200, which is less than 100,000, the billable CU count for that hour is 100,000.

• If the actual number of CUs used in an hour is 150,200, which is greater than 100,000, the final billable CU count for that hour is 151,000. This is because the minimum billing granularity is 1,000 CUs.

CU measurement rules for different APIs

CUs per API call depend on the API type coefficient and payload size:

CUs for a single API call = CU coefficient × CEILING{(Request payload + Response payload) / 512 KB}

CU coefficients by API type:

• Model API: 10

• MCP Server: 5

• Agent API: 2

Billing examples:

1. For a Model API request with a 100 KB request payload and an 800 KB response payload, the CUs for this request are calculated as: 10 × CEILING{(100 + 800) / 512} = 20

2. For an MCP Server request with a 50 KB request payload and a 300 KB response payload, the CUs for this request are calculated as: 5 × CEILING{(50 + 300) / 512} = 5

CU price