Doc intelligence model not found

Hello @INFOSOFT

Thank you for contacting Microsoft Q&A and for sharing the detailed troubleshooting information.

Based on the behavior described, the issue appears to be related to the custom model being associated with a different Azure Document Intelligence resource than the endpoint currently being queried through the REST API.

From your observations:

• The custom model training status is Succeeded

The model works correctly within Document Intelligence Studio

GET /documentintelligence/documentModels?api-version=2024-11-30 returns only prebuilt models

Requests to the custom model return 404 ModelNotFound

Older /formrecognizer/... endpoints return 404 ResourceNotFound

this indicates that the endpoint itself is reachable and functioning correctly, but the specific resource being queried does not currently contain any registered custom models.

Please review the following validation steps:

1. Verify the exact Azure Document Intelligence resource

In Azure AI Foundry / Document Intelligence Studio:

• Open the project settwings

Identify the connected Azure Document Intelligence resource

Confirm:

Resource name

  Subscription
  
     Region
     

Then compare this against the REST endpoint currently being used.

The endpoint should match the resource exactly in the following format:

https://<resource-name>.cognitiveservices.azure.com

Please note that no explicit regional prefix (such as centralindia) is required in the URL, as the region is already embedded in the resource DNS.

2. Validate the API path and API version

For Document Intelligence v4.0 GA, please use the following endpoint format consistently:

GET https://<resource-name>.cognitiveservices.azure.com/documentintelligence/documentModels?api-version=2024-11-30

We recommend avoiding mixing:

/formrecognizer/... and

/documentintelligence/...

across different API versions, as this may lead to 404 responses depending on the model generation and endpoint compatibility.

3. Validate the custom model directly

Please retrieve the exact custom model ID from Studio and test the following request:

GET https://<resource-name>.cognitiveservices.azure.com/documentintelligence/documentModels/{modelId}?api-version=2024-11-30

Expected behavior:

• If the model exists on that resource, model metadata should be returned
• If the request returns 404 ModelNotFound, it confirms the model is not associated with that specific resource endpoint

4. Validate authentication and network configuration

Please also verify:

The API key being used belongs to the same Document Intelligence resource

No firewall, VNet, or private endpoint restrictions are blocking the request

The request is targeting the actual Cognitive Services endpoint and not:

• AI Foundry project endpoint
• Hub endpoint
• Agent endpoint
• Inference endpoint

Since the /documentModels API successfully returns prebuilt models and no pagination is present, this strongly suggests:

The service endpoint is healthy

Authentication is functioning

But no custom models are currently registered under that resource

At this stage, the most likely root cause is a mismatch between:

• The resource connected in Studio and
• The resource endpoint being queried via REST API

Please refer this

Document Models REST reference (list & get) https://learn.microsoft.com/rest/api/aiservices/document-models/list-models?view=rest-aiservices-v4.0

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

Thank you!

i think GET /documentModels lists the models that exist in that DI resource, and a custom model appears there when it is created in that same resource in the Portal.

Hello INFOSOFT,

Greetings! Thanks for raising this question in Q&A forum.

The reason your custom trained model works fine in Document Intelligence Studio but does not appear when you call the documentModels list API is almost certainly an endpoint mismatch or an API version compatibility issue. Document Intelligence Studio sometimes connects to your resource using its own internal session context, which can mask the fact that your API call is hitting a slightly different endpoint, resource, or using an API version that does not return custom models the same way. This is a very common and easy-to-fix issue.

Here are the steps to identify and resolve this quickly:

Step 1: Verify you are calling the correct resource endpoint

This is the most common cause. Make sure the endpoint in your API call exactly matches the resource endpoint shown in the Azure Portal — not the Studio URL. Go to:

Azure Portal → Your Document Intelligence Resource → Keys and Endpoint → Copy the Endpoint value

It should look like:

https://<your-resource-name>.cognitiveservices.azure.com/

Your API call should then be:

GET https://<your-resource-name>.cognitiveservices.azure.com/documentintelligence/documentModels?api-version=2024-11-30

If you are using a different base URL (for example, one with a region prefix like centralindia.api.cognitive.microsoft.com), that older endpoint format may not return custom models correctly with the newer API version.

Step 2: Try calling the model directly by its model ID

Instead of listing all models, try fetching your specific custom model directly by its ID to confirm the API can see it:

GET https://<your-resource-name>.cognitiveservices.azure.com/documentintelligence/documentModels/<your-model-id>?api-version=2024-11-30

If this returns your model details successfully, the model exists and is accessible — the issue is only with how the list call filters or paginates results.

Step 3: Check for pagination in the list response

The documentModels list API paginates results and pre-built models typically appear first. Your custom model may be on the next page. Check the response body for a nextLink field:

{
  "value": [...],
  "nextLink": "https://<endpoint>/documentintelligence/documentModels?api-version=2024-11-30&$skipToken=..."
}

If nextLink is present, follow that URL to retrieve the next page — your custom model will likely appear there.

Step 4: Confirm you are using the correct API key

If you have multiple Document Intelligence resources in the same region, it is easy to accidentally list models on the wrong resource. Double-check that the API key you are sending in the Ocp-Apim-Subscription-Key header belongs to the same resource where you trained the custom model. Go to:

Azure Portal → Your Document Intelligence Resource → Keys and Endpoint → confirm Key 1 or Key 2 matches what you are using in the API call.

Step 5: Try the older API version as a cross-check

If 2024-11-30 is not returning your custom model, try the same list call with an earlier stable API version to see if the model appears:

GET https://<your-resource-name>.cognitiveservices.azure.com/documentintelligence/documentModels?api-version=2023-07-31

If your model shows up with the older version but not with 2024-11-30, this points to a regional API version rollout issue in centralindia and should be reported to Microsoft support.

Step 6: Open an Azure Support Ticket if the model is still missing

If you have confirmed the correct endpoint, API key, model ID, and pagination but the custom model still does not appear via the API while working fine in Studio, this is a backend inconsistency that needs Microsoft investigation. Go to:

Azure Portal → Help + Support → New Support Request and fill in:

• Issue type: Technical
• Service: Azure Document Intelligence
• Problem type: Custom Model Issues
• Problem subtype: Model not returned in API list
• Severity: B (Moderate)

In the description include:

• Your resource name and region (Central India)
• Your custom model ID and its training status (Succeeded)
• The exact API call you are making with the API version
• Confirmation that the model works in Studio but not via the REST API
• Steps you have already tried from this list

If this answer helps you kindly accept the answer which will help others who have similar questions.

Best Regards,

Jerald Felix.