Microsoft-Decision-1 returns HTTP 400 unknown_model in Southeast Asia despite successful deployment

Pace Port 25 0 Reputation points
2026-10-20T03:21:13.5566667+00:00

Hi Microsoft Team,

I'm experiencing an inference issue with Microsoft-Decision-1 (Preview) in Microsoft Foundry, Southeast Asia.

Deployment configuration:

  • Region: Southeast Asia

Resource type: Azure AI Services (AIServices)

Model: Microsoft-Decision-1

Model version: 1

Deployment name: Decision-1

Deployment type: DataZoneStandard

Provisioning state: Succeeded

The model appears in the Southeast Asia regional catalog with decision: true, and DataZoneStandard is listed as a supported deployment type.

Issue

Although the deployment succeeds, inference requests to the documented Decision API endpoint consistently fail:

POST https://<resource-endpoint>/providers/microsoft/v1/systemone

The request body uses the correct deployment name:

{
  "model": "Decision-1",
  "state": "The user wants to summarise a short email.",
  "questions": {
    "model_tier": {
      "type": "choice",
      "instructions": "Select the most appropriate model tier.",
      "criteria": {
        "efficient": "Simple summarisation and extraction tasks",
        "capable": "Complex reasoning and multi-step analysis"
      }
    }
  }
}

Error response:

{
  "error": {
    "code": "unknown_model",
    "message": "Unknown model: Microsoft-Decision-1",
    "details": "Unknown model: Microsoft-Decision-1"
  }
}

Troubleshooting performed:

Confirmed deployment provisioning state is Succeeded.

Verified that Microsoft-Decision-1 version 1 is listed in the Southeast Asia model catalog.

Confirmed that the request uses the deployment name (Decision-1), not the underlying model name.

Reproduced the error using both Python and a direct curl request.

Successfully tested Microsoft-Decision-1 in Sweden Central, suggesting a possible region-specific issue.

Our project has APAC data-residency requirements, so Sweden Central is not an acceptable workaround.

Questions:

Is Microsoft-Decision-1 inference currently operational in Southeast Asia?

Is there a known issue with the preview model's inference backend or deployment registration in this region?

Does the model require additional subscription-level access or feature enablement?

Is there an expected resolution or a supported APAC-region workaround?

Any guidance from the Microsoft Foundry team would be greatly appreciated.

Thank you.Hi Microsoft Team,

I'm experiencing an inference issue with Microsoft-Decision-1 (Preview) in Microsoft Foundry, Southeast Asia.

Deployment configuration:

Region: Southeast Asia

Resource type: Azure AI Services (AIServices)

Model: Microsoft-Decision-1

Model version: 1

Deployment name: Decision-1

Deployment type: DataZoneStandard

Provisioning state: Succeeded

The model appears in the Southeast Asia regional catalog with decision: true, and DataZoneStandard is listed as a supported deployment type.

Issue

Although the deployment succeeds, inference requests to the documented Decision API endpoint consistently fail:

POST https://<resource-endpoint>/providers/microsoft/v1/systemone

The request body uses the correct deployment name:

{
  "model": "Decision-1",
  "state": "The user wants to summarise a short email.",
  "questions": {
    "model_tier": {
      "type": "choice",
      "instructions": "Select the most appropriate model tier.",
      "criteria": {
        "efficient": "Simple summarisation and extraction tasks",
        "capable": "Complex reasoning and multi-step analysis"
      }
    }
  }
}

Error response:

{
  "error": {
    "code": "unknown_model",
    "message": "Unknown model: Microsoft-Decision-1",
    "details": "Unknown model: Microsoft-Decision-1"
  }
}

Troubleshooting performed:

Confirmed deployment provisioning state is Succeeded.

Verified that Microsoft-Decision-1 version 1 is listed in the Southeast Asia model catalog.

Confirmed that the request uses the deployment name (Decision-1), not the underlying model name.

Reproduced the error using both Python and a direct curl request.

Successfully tested Microsoft-Decision-1 in Sweden Central, suggesting a possible region-specific issue.

Our project has APAC data-residency requirements, so Sweden Central is not an acceptable workaround.

Questions:

Is Microsoft-Decision-1 inference currently operational in Southeast Asia?

Is there a known issue with the preview model's inference backend or deployment registration in this region?

Does the model require additional subscription-level access or feature enablement?

Is there an expected resolution or a supported APAC-region workaround?

Any guidance from the Microsoft Foundry team would be greatly appreciated.

Thank you.

Microsoft Foundry
Microsoft Foundry

A unified Azure platform for creating and managing AI models, agents, and applications with built‑in enterprise security, monitoring, and governance

0 comments

1 answer

Sort by: Most helpful
Most helpful Newest Oldest

  1. AI answer

    2026-10-20T03:22:10.2933333+00:00

    The documented configuration is supported in Southeast Asia, so a consistently reproducible unknown_model response from a succeeded deployment should be escalated as a regional backend/deployment-registration failure rather than addressed by changing the request payload.

    What is confirmed

    • Microsoft-Decision-1 version 1 supports DataZoneStandard in Southeast Asia. The same APAC matrix also lists Australia East, Japan East, Japan West, Korea Central, and South India.^1^
    • Your endpoint is correct: 
        <resource-endpoint>/providers/microsoft/v1/systemone
    • The request must place the deployment name—Decision-1 in your case—in the model property. Your payload follows that contract.^2^
    • DataZoneStandard keeps inference processing within the APAC data zone, although processing is not guaranteed to remain specifically in Southeast Asia.^1^

    No public Microsoft documentation cited here identifies unknown_model on this endpoint as a currently declared Southeast Asia incident or provides an estimated resolution time. A related Q&A report describes catalog/deployment availability inconsistencies, but it has no accepted or recommended resolution.^3^

    Required next action

    Open the affected Azure AI Services resource in the Azure portal, then select:

    1. Help → Support + Troubleshooting.
    2. Run the offered diagnostics for:
      • Problem related to a region
      • Model Deployments
    3. If unresolved, select Create a support request.
    4. Classify it as a technical problem involving Microsoft Foundry model inference and regional deployment.
    5. Include:
      • Resource region: southeastasia
      • Resource kind: AIServices
      • Model/version: Microsoft-Decision-1 / 1
      • Deployment/SKU: Decision-1 / DataZoneStandard
      • Deployment provisioning state
      • UTC timestamps of failed requests
      • Full response headers and any request, correlation, or APIM request IDs
      • The sanitized request and response
      • Confirmation that the identical operation succeeds in Sweden Central
      • A screenshot/export showing the deployment details

    Azure documents this resource-scoped support path and allows diagnostic files or traces to be attached.^4^

    Do not add automatic retries for this response: HTTP 400 is not a transient status. Retrying the same deployment and payload will not repair missing backend registration.

    Access and quota

    The documented prerequisites are a paid Azure subscription, Foundry access, deployment-management permission, and either Microsoft Entra ID or API-key authentication. Cognitive Services Contributor is an example role that permits model deployment.^2^

    The documentation does not specify an additional subscription-level preview enrollment for this model. Because your deployment reaches Succeeded and the service returns a model-resolution error—not an authorization error—there is no supported basis for treating extra access enablement as the primary fix.

    APAC-resident workaround

    If APAC-wide processing satisfies your residency policy, create a separate DataZoneStandard deployment in another listed APAC resource region—Australia East, Japan East/West, Korea Central, or South India—and test its own copied resource endpoint. DataZoneStandard processes inference within APAC.^1^

    Do not switch to GlobalStandard when APAC-only inference is mandatory: global deployments may process requests in any Azure region.^5^

    References

    1. Region availability for Foundry Models sold by Azure - Microsoft ...
    2. Deploy and use Microsoft-Decision-1 in Microsoft Foundry
    3. AI Foundry Microsoft-Decision-1 listed as available in all regions, but ...
    4. Create an Azure support request
    5. Deployment types for Microsoft Foundry Models
    AI-generated content may be incorrect. Read our transparency notes for more information.

    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.