[Vertex AI Bug] Project cannot access ANY Gemini models ("Listed 0 items") - Extensive Troubleshooting Done

Google Cloud Build with AI Custom ML & MLOps
, , ,
1

Hello everyone, Google team, Hoping someone can help me out here because this is driving me crazy.

I’m completely blocked from using any Gemini models through Vertex AI on my project, and I’ve tried pretty much everything. Here’s the problem: When I run `gcloud ai models list --region=us-central1 --project=(PII Removed by Staff)l`, I just get `Listed 0 items.`. Every single time. Because of this, trying to call any Gemini model (like `gemini-1.0-pro` or `gemini-1.5-flash-latest`) from my Cloud Function fails with the “404 Not Found: Publisher Model was not found or your project does not have access to it” error. But here’s the weird part: If I go to Google AI Studio using the exact same Google account and project (PII Removed by Staff), it works perfectly! I can chat with Gemini models like `gemini-1.5-flash-latest` no problem.

So my account does have permission to use Gemini, right? Here’s what I’ve already tried (and failed):

  • Billing: My account is fully active (paid account, NOT free trial, for over a month) with a good credit card. No billing alerts.

  • Vertex AI API: Yes, `aiplatform.googleapis.com` API is enabled.

  • New Project: I even created a completely fresh project (PII Removed by Staff) thinking my old one was cursed. Same exact problem on the new one.

  • Regions: Tried deploying and running the gcloud command in both `us-central1` and `europe-west1`. Still `Listed 0 items.`.

  • Permissions: Checked IAM roles. Gave `Vertex AI User` to the default service account, even tried a dedicated one. Didn’t help.

  • Cloud Functions: Deployed my Python code using both the official `google-cloud-aiplatform` library (fails with the 404 error) AND tried bypassing Vertex using the `google-generativeai` library with a fresh AI Studio API key. That also fails inside the Cloud Function (weird `v1beta` error in the logs: `404 models/… is not found for API version v1beta…`), even though the same key works perfectly in AI Studio directly.

So, what gives? It really feels like there’s some kind of provisioning bug specifically with the Vertex AI platform for my project (PII Removed by Staff). It just refuses to see the models, even though AI Studio proves my account can access them.

Could someone from the Google Cloud / Vertex AI team please take a look at why Vertex AI isn’t listing the models for project (PII Removed by Staff)?

I’m totally blocked here.

Thanks a lot for any help. (pleeeease)

2

Hello @75_020,

I think you should check this page: Which Gemini model should I migrate to?

Gemini 1.5 models are retired models, so GCP projects should no longer be able to use them.

Can you share a screenshot from AI Studio to show the configuration of the chat? It’s either a graphical bug, you’re not using the correct model or project, or it’s just a bug in AI Studio.

Regarding your GCP Project, since models are retired, you need to migrate to a newer one, otherwise, your cloud function won’t be able to find any models that are now retired.