When deploying ML model to online endpoint are you charged for the time the endpoint is up or the time it is running?

Question

If I deploy ML model to online endpoint say for 1 day but it is invoked only once and runs for 1min, then am I charged for the whole day or for the 1min?

Answer 1

@Anonymous charges for the endpoint are based on the compute that is used for the endpoint. If the compute is up and running the charges will be applicable along with networking charges for the traffic that is generated when requests are made.

See the note on this page for section Manged Online endpoints which explains how the charges are incurred. It is recommended to lookup costs of your endpoint through cost analysis page of your subscription. This allows you to monitor the costs regularly and moderate the usage of endpoint when it is not required.

To summarize, you will be charged for compute that is up and running for online endpoints.

I have answered couple of similar questions for another user before, Please see these threads for more information on online & batch endpoints.

https://learn.microsoft.com/en-us/answers/questions/1165606/cost-for-online-endpoint-in-azure-machine-learning

https://learn.microsoft.com/en-us/answers/questions/1182960/cost-of-batch-endpoint-in-azure-machine-learning

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.