Gemini 3.1 Flash Image started billing at unsupported "Priority PayGo" SKU (Cost doubling by surprise)

Google Cloud Build with AI AI APIs
1

Is anyone else seeing unexpected billing spikes for gemini-3.1-flash-image starting on March 27?

My project has a completely frozen codebase (using google-cloud-aiplatform==1.133.0, no deployments since March 19). However, starting on March 27, all of my standard image generation traffic silently migrated on the backend from the Standard SKU (7D83-2D44-B50E) to the Priority SKU (EC3C-2188-6892), effectively doubling my costs.

The catch: According to the official Vertex AI model documentation for Gemini 3.1 Flash Image, Priority PayGo is explicitly NOT supported for this model.

Because Priority isn’t supported, it’s impossible for my code to have requested it. It appears the Vertex AI global load balancers are currently defaulting unflagged requests for this preview model to a non-existent premium tier and passing the cost to users.

Has anyone else caught this in their billing report?

4

You should definitely create a billing support ticket for this. yes Priority is not supported method for the image generation models. Vertex AI Pricing  |  Google Cloud

5

I’m facing same issue, I’m using gemini-flash-2.5 and sending no headers specifically to run priority, still the costs spiked alost 2x suddenly from 27th march, its visible in reports that the runs are happening using priority input/output.

6

I also see “priority” across multiple models appearing out of nowhere on Mar 27 and then disappearing today (Mar 31). Writing support ticket now to get a refund.

7

Indeed, it was solved today. Maybe due to an investigation they started after I pointed it out two days ago.