Share via

Missing model-router deployment in Azure AI Foundry

Luca L 20 Reputation points
2025-09-01T07:01:43.0533333+00:00

Hi,
We are trying to test the integration of model-router in our applications. However, it does not appear in our model catalog - see screenshot below from a Sweden Central resource, even though we already have valid o3 access.

According to the documentation, no additional request should be required for model-router access (“If you already have o3 access, no request is required.”).

We also verified our Azure OpenAI services in both East US 2 and Sweden Central, where we already have some Global Standard deployments, but the model-router option is still missing.

Could you please advise why it might not be available in our catalog, or let us know if there are any extra steps needed to enable it?

Thanks in advance for your support!

User's image

Azure OpenAI in Foundry Models

Answer accepted by question author

SRILAKSHMI C 19,090 Reputation points Microsoft External Staff Moderator
2025-10-01T18:01:36.3533333+00:00

Hi Luca L,

I sincerely apologize for the delayed response.

Thank you for the detailed follow-up and for sharing your observations this is extremely helpful.

You’re absolutely right in your assessment. Based on similar cases we’ve seen, the behavior you’re describing typically occurs when older Azure OpenAI resources haven’t fully synchronized with the latest catalog metadata for limited-access features like model-router. This is why the option might not surface in the portal UI for existing instances, even though deployments via Terraform (or freshly created resources) work as expected.

At this time, there’s no functional impact on your deployments as you’ve confirmed, the model-router works correctly once provisioned, regardless of whether it was created through Terraform or a new instance. The issue is primarily related to the portal UI not reflecting the most recent feature availability on older resources.

In the meantime, your suggested workarounds using Terraform or deploying via a newly created Azure OpenAI resource are the recommended approaches. We appreciate you highlighting them here for the benefit of others who might face the same issue.

Thank you!

Was this answer helpful?

1 person found this answer helpful.

2 additional answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. Luca L 20 Reputation points
    2025-09-15T07:54:54.5533333+00:00

    We’ve been able to create a model-router deployment successfully from the UI, but only when using a freshly created Azure OpenAI instance. On our older instances, the option is still missing from the catalog.

    Interestingly, when deploying via Terraform, everything works fine across both old and new instances. So this seems to be more of a UI glitch rather than an actual access/configuration issue.

    Sharing here in case others run into the same behavior: trying with a new instance or using Terraform can help as a workaround.

    Was this answer helpful?

  2. Amira Bedhiafi 42,846 Reputation points MVP Volunteer Moderator
    2025-09-01T19:47:03.6066667+00:00

    Hello Luca !

    Thank you for posting on Microsoft Learn Q&A.

    Model-router is still a limited-access preview and only shows up under specific conditions. Even though the doc says "if you already have o3 access, no request is required" Microsoft still gates the latest model-router version and only surfaces it in certain regions & deployment types. https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/model-router

    Use Azure AI Foundry, in your project then models and create deployment (wizard). Many tenants don’t see model-router on the explore grid, but it does appear in the deployment picker when eligible.

    Then clear catalog filters (providers = all, access = all, availability = all) and make sure you’re in the subscription/tenant that has your OpenAI resource.

    Model-router is only listed for Global Standard deployments, in East US 2 and Sweden Central right now. If your Project is pinned to another deployment type (Data zone / Provisioned) or another region, it won’t show. Create the deployment as Global Standard in one of those two regions.

    I check the the link and it says that o3 access is sufficient, but in practice the UI hides it if your org hasn’t been flipped for limited accesson that version. If it still doesn’t appear in the deployment picker, you can submit the GPT-5 limited-access form or open a support ticket to enable model-router for your tenant.

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.