A lot of cloud services sorta work the same way. AWS and Azure are pay per request for all sorts of things, I figured that was the model the inference providers were following.