Why are custom models visible in Studio but not accessible via API

Question

There appears to be a disconnect between Document Intelligence Studio and the actual Document Intelligence resource. Custom models visible in Studio are not accessible via the API and do not appear when listing models programmatically. Additionally, Project Settings cannot be saved in Studio - when attempting to link the project to a specific resource by entering endpoint and key, the "Save" button does not persist the changes.

Answer 1

Hello Noya Gendelman,

Welcome to Microsoft Q&A.

Thanks for reporting this. What you’re seeing is confusing, but it usually comes down to how Document Intelligence Studio projects are linked to actual Azure resources. I’ll explain what’s happening and how to resolve it.

Why models are visible in Studio but not accessible via the API

Document Intelligence Studio can display project-level custom models even when those models are not actually registered in your Azure Document Intelligence resource.

In simple terms:

Studio allows you to create projects, label data, and even train models locally within the Studio experience.

A model becomes accessible via REST/SDK only after it is successfully trained and published to a specific Document Intelligence resource.

If the project is not properly linked to a resource, Studio may still show the models, but they do not exist from the service/API perspective.

This is why:

• The models don’t appear when you list models programmatically.
• API calls fail to find or use them.

Why Project Settings are not saving

When the endpoint and key do not persist after clicking Save, it indicates that Studio is not successfully binding the project to the Azure resource. Common causes include:

The endpoint or key belongs to a different region than the Studio project

The key is from a different subscription or resource

Your user account does not have Contributor (or higher) permissions on the Document Intelligence resource

Browser or session issues (cached data, blocked cookies, private/incognito mode)

When this happens, Studio silently keeps the project in a detached state, even though the UI still allows you to work with it.

What this means in practice

Right now:

The models you see exist only in Studio project metadata

The Document Intelligence service itself does not know about them

The API behavior is correct in not returning models that were never published to the resource

This is not a billing or service outage issue.

Here are some troubleshooting steps,

Verify the resource

• Confirm the endpoint and key are from the correct Document Intelligence resource
• Ensure the resource region matches the Studio project region

Check permissions

• Make sure your user has at least Contributor access on the resource (check IAM in Azure Portal)

Re-link the project

• Open Project Settings in Studio
• Re-enter the endpoint and key
• Save, then refresh the page (try a different browser if needed)

Retrain and publish

• Retrain the custom model after the project is correctly linked
• Confirm the model appears via:

    GET /documentModels
  

Validate via API

• Test using REST, SDK, or Postman to confirm the model is listed and callable.

This is not a model failure. It’s a resource binding issue between Document Intelligence Studio and the underlying service.

Once the project is properly linked and the model is retrained, Studio and the API will align and the models will be accessible programmatically.

Please refer this

• Build and Train Azure Document Custom Models Using API or SDK
• Document Intelligence custom models

I Hope this helps. Do let me know if you have any further queries.

Thank you!