Help with Vertex AI: 404 Not Found on Gemini ModelsDespite Correct IAM Permissions (new project)

[maurosg78]

Hello everyone,

I'm running into a persistent access issue with Vertex AI on a new project and would be very grateful for any insights.

The Problem: When calling any Gemini model (e.g., gemini-1.5-pro, gemini-1.0-pro) from a Node.js Cloud Function using its service account, I consistently receive a 404 Not Found error with the message: Publisher Model ... was not found or your project does not have access to it.

Troubleshooting Steps Already Taken:

To save time, here is a complete list of what I have already verified:

• API Enabled: The Vertex AI API (aiplatform.googleapis.com) is enabled for the project.
• Billing Active: An active billing account is correctly linked to the project.
• IAM Permissions: The service account has the required Vertex AI User role. My personal user account is the project Owner.
• Multi-Region Tests: The 404 error persists across multiple regions, including us-central1 and us-east1.
• Manual Console Test: I can successfully call the exact same Gemini models from the Vertex AI Studio (the web console playground) when logged in with my personal user account. The issue is strictly limited to programmatic access via the service account.
• Project "Warm-up": We ran a script for over 60 hours making continuous, successful calls to another Google Cloud API (Cloud Translation) to build a history of legitimate use. This had no effect.
• gcloud CLI Diagnostics: We have confirmed via gcloud that there are no restrictive Organization Policies applied to this project.

My Current Hypothesis: After exhausting all technical diagnostics, my only conclusion is that there is an administrative or sandbox-style restriction on new projects that prevents service accounts from accessing Gemini models programmatically, even if a human user can access them via the console.

My Questions for the Community:

1. Has anyone else experienced this specific issue where a service account is blocked from Vertex AI on a new project, while a user account is not?
2. Is there a manual activation step, a waitlist, or a specific Terms of Service acceptance required for programmatic access that might be missing?
3. Given that our project is on the basic (billing-only) support plan, is there any known path to request this access be enabled?

Any ideas or suggestions would be greatly appreciated.

Thank you in advance.

[marckevin · Accepted solution]

Hi @maurosg78,

Welcome to Google Cloud Community!

Please see my answers inline with your questions below:

1. Has anyone else experienced this specific issue where a service account is blocked from Vertex AI on a new project, while a user account is not?

   Yes, this is possible. It is a common challenge when dealing with 404 errors, especially with new projects. Additionally, a 404 error when calling the Gemini models through the Vertex AI API may occur if you're using a retired or legacy model. Legacy models are restricted from new projects, and API requests to retired models will fail. Ensure that you are not calling a legacy or retired Gemini model.

2. Is there a manual activation step, a wait list, or a specific Terms of Service acceptance required for programmatic access that might be missing?

   A wait list form, or a specific Terms of Service acceptance required for programmatic access depends on the model you are accessing. Usually models under private preview. To check model versions and lifecycle, you can refer to this documentation.

3. Given that our project is on the basic (billing-only) support plan, is there any known path to request this access be enabled?

   For complete reference on the different support plans and their coverage, you can refer to this documentation.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.

[marckevin · Accepted solution]

Hi @maurosg78,

Welcome to Google Cloud Community!

Please see my answers inline with your questions below:

1. Has anyone else experienced this specific issue where a service account is blocked from Vertex AI on a new project, while a user account is not?

   Yes, this is possible. It is a common challenge when dealing with 404 errors, especially with new projects. Additionally, a 404 error when calling the Gemini models through the Vertex AI API may occur if you're using a retired or legacy model. Legacy models are restricted from new projects, and API requests to retired models will fail. Ensure that you are not calling a legacy or retired Gemini model.

2. Is there a manual activation step, a wait list, or a specific Terms of Service acceptance required for programmatic access that might be missing?

   A wait list form, or a specific Terms of Service acceptance required for programmatic access depends on the model you are accessing. Usually models under private preview. To check model versions and lifecycle, you can refer to this documentation.

3. Given that our project is on the basic (billing-only) support plan, is there any known path to request this access be enabled?

   For complete reference on the different support plans and their coverage, you can refer to this documentation.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.

[ericdong · Accepted solution]

@maurosg78 : Gemini 1.0/1.5 models (e.g. gemini-1.5-pro, gemini-1.0-pro) are not longer available for new users/projects. Please see this page for the latest available models: https://cloud.google.com/vertex-ai/generative-ai/docs/models