Azure Open AI (Not Foundry)

Question

Azure Open AI (Not Foundry)

Operations 0

I have created an Azure Open AI resource (not the Foundry option) however I cant work out how to call the api for the specific models e.g. GPT‑4o‑mini. There does not seem to be any option to specify or deploy a model when creating the Azure Open AI resource or from the Created resource within the Portal ?

Sai Kiran Mudavath 90 Reputation points Microsoft External Staff Moderator

2026-10-01T09:27:12.85+00:00
Hi @Operations ,

An Azure OpenAI resource provides the service endpoint, but a model such as gpt-4o-mini must be deployed separately before you can use it for inference.

Since you created an Azure OpenAI resource rather than a new Foundry resource, you can deploy the model through the Foundry classic portal:

Sign in to Microsoft Foundry at https://ai.azure.com and make sure the New Foundry toggle is off.

Under Keep building with Foundry, select View all resources.

Select your Azure OpenAI resource. If prompted to upgrade the resource to Foundry, select Cancel to continue with the existing Azure OpenAI resource.

In the left pane, under Shared resources, select Deployments.

Select + Deploy model → Deploy base model.

Select the required model, such as gpt-4o-mini, and select Confirm.

Configure the Deployment name and Deployment type, then select Deploy.

After the deployment is created, use the deployment name when making API calls rather than the underlying model name.

For the complete steps, please refer to Create and deploy an Azure OpenAI in Microsoft Foundry Models resource (classic).

1 answer

Your answer

Sai Kiran Mudavath 90 Reputation points Microsoft External Staff Moderator

2026-10-01T09:27:12.85+00:00

Hi @Operations ,

An Azure OpenAI resource provides the service endpoint, but a model such as gpt-4o-mini must be deployed separately before you can use it for inference.

Since you created an Azure OpenAI resource rather than a new Foundry resource, you can deploy the model through the Foundry classic portal:

Sign in to Microsoft Foundry at https://ai.azure.com and make sure the New Foundry toggle is off.

Under Keep building with Foundry, select View all resources.

Select your Azure OpenAI resource. If prompted to upgrade the resource to Foundry, select Cancel to continue with the existing Azure OpenAI resource.

In the left pane, under Shared resources, select Deployments.

Select + Deploy model → Deploy base model.

Select the required model, such as gpt-4o-mini, and select Confirm.

Configure the Deployment name and Deployment type, then select Deploy.

After the deployment is created, use the deployment name when making API calls rather than the underlying model name.

For the complete steps, please refer to Create and deploy an Azure OpenAI in Microsoft Foundry Models resource (classic).

Answer 1

An Azure OpenAI resource is only the service endpoint. It does not automatically select a model for requests.

To use a model such as GPT-4o or GPT-4o-mini, a model deployment must be created after the resource exists.

Supported programmatic model names include:

gpt-4o version 2024-11-20
gpt-4o version 2024-08-06
gpt-4o version 2024-05-13
gpt-4o-mini version 2024-07-18

A working flow is:

Create the Azure OpenAI resource in a supported standard or global-standard region.
Deploy the model to that resource.
Call the API against the deployment name created for that model.

In the portal flow shown in the documentation:

Open Azure AI Foundry and sign in with the credentials associated with the Azure OpenAI resource.
Select the correct directory, subscription, and Azure OpenAI resource.
Go to Models + endpoints.
Select + Deploy model.
Choose gpt-4o or gpt-4o-mini.
Enter a deployment name and deploy it.
Use that deployment in the playground or API calls.

Important detail: the endpoint and key belong to the Azure OpenAI resource, but the model used by an API call is determined by the deployment name, not just by the resource endpoint. If multiple models are deployed under one resource, the request must specify the intended deployment.

If deployment is being done programmatically instead of through the portal, the documented model names above are the names to use when creating the deployment.