500 Server Error When Requesting Ephemeral Token for gpt-40-realtime-preview on Sweden Central

Question

500 Server Error When Requesting Ephemeral Token for gpt-40-realtime-preview on Sweden Central

Cristina S 5

Hello,

I am currently trying to fetch an ephemeral API token for the REALTIME API using gpt-40-realtime-preview in the Sweden Central region. However, I keep receiving a 500 Server Error, even when using the boilerplate example from the documentation (with my own api_key and specifying the model as the deployment name).

Here is the error message:"message": "The server had an error processing your request. Sorry about that! You can retry your request, or contact us by filing an Azure support request in the portal. You can find information on how to do that here: https://learn.microsoft.com/azure/ai-services/cognitive-services-support-options if you keep seeing this error. (Please include the request ID 92b5e5ec-c119-4e17-8200-4fd571f999c0 in your email.)",

1 answer

Your answer

Answer 1

Jerald Felix 1,705

Hello Cristina S,

This typically points to a backend issue on Azure’s side, not a problem with your API key or payload syntax.

Double-check a few critical areas:

Ensure gpt-40-realtime-preview is the actual name of your deployment in Azure, not just the model name.

Must be: https://<your-resource-name>.openai.azure.com/openai/deployments/<deployment-name>/chat/completions?api-version=2024-02-15-preview

Confirm that gpt-40-realtime-preview is officially available in Sweden Central. As of mid-May 2025, this model may not be live in that region yet.

Azure has been rolling out GPT-4o (gpt-40-realtime-preview) gradually, and some regions are still in preview or may be experiencing instability.

Visit the official Azure OpenAI Status Page to check if Sweden Central or GPT-4o services are down.
Review the GPT-4o rollout announcement or Azure updates page for Sweden Central support status.

Best Regards,
Jerald Felix

Cristina S 5 Reputation points

2025-05-19T19:18:01.55+00:00

Hi Jerald Felix,

Thank you for your answer. I am trying to implement a speech-to-speech functionality using realtime-preview and WebRTC, following the documentation: https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/realtime-audio-webrtc

The model is correctly deployed because I can use Realtime speech-to-speech in the Foundry playground.

As stated in the documentation updated in April, the ephemeral API key should be retrieved using this URL example:

https://YourAzureOpenAIResourceName.openai.azure.com/openai/realtimeapi/sessions?api-version=2025-04-01-preview

It seems my request has the correct body and parameters because if I change either of them, I get 404 'Resource not found.

After I make the request, this shows up in the Resource health monitoring:

Unavailable (Unplanned) : Resource health event (Unplanned)

We are sorry, your Cognitive Services resource is unavailable. We're working to automatically recover your Cognitive Services resource and to determine the source of the problem. No additional action is required from you at this time.

It fixes itself after some time and on the next request does the same.
Cristina S 5 Reputation points

2025-05-19T20:28:58.1+00:00

Solved it, the request body was missing the parameter "modalities" : ["audio","text"]. It wasn't present in the WebRTC example nor was it listed as required in the documentation.

Share via

500 Server Error When Requesting Ephemeral Token for gpt-40-realtime-preview on Sweden Central

1 answer

Your answer