Azure OpenAI

Add and configure Azure OpenAI models through TrueFoundry’s AI Gateway.

Adding Models

This section explains the steps to add Azure OpenAI models and configure the required access controls.

1. Navigate to Azure OpenAI Models in AI Gateway

   From the TrueFoundry dashboard, navigate to AI Gateway > Models and select Azure OpenAI.

   

   Navigate to Azure OpenAI Models

2. Add Azure OpenAI Account Details

   Click Add Azure OpenAI Account. Give a unique name to your Azure OpenAI account and complete the form with your Azure authentication details (API Key). For enhanced security, you can also use certificate-based authentication, client secret based authentication, or workload identity federation. Add collaborators to your account. You can read more about access control here.

   
3. Add Models by Deployment Name

   Click on + Add Model to open the form for adding a new model. For Azure OpenAI, you don’t select from a list. Instead, you add models based on your deployments in Azure. The Model ID in TrueFoundry corresponds to your Deployment Name in Azure. First, ensure you have deployed a model in your Azure OpenAI service. You can follow Microsoft’s instructions here. Once deployed, you can find the deployment name in the Deployments tab of your Azure AI Studio.

   

   Find Model Deployment Name in Azure

   Use this deployment name as the Model ID when adding the model in TrueFoundry.

   

   Adding an Azure OpenAI model to TrueFoundry

   Azure OpenAI integration supports any OpenAI model that you have deployed in your Azure account.

Reasoning models

Azure exposes a model’s reasoning only through the Responses API, so when a Chat Completions request asks for reasoning — via reasoning_effort, a thinking field, or the x-tfy-openai-use-responses-api header — the AI Gateway serves it through Azure’s Responses API upstream and converts the reply back. Your client stays on the Chat Completions contract.

Two Azure-specific conditions apply:

• The deployment’s api-version must be 2025-03-01-preview or newer. Older versions reject /responses outright, so the AI Gateway keeps those requests on Chat Completions and no reasoning is returned.
• The deployment’s foundation model must be Responses-capable in the model catalogue. Azure deployment names are arbitrary, so the AI Gateway matches on the underlying foundation model rather than the deployment name you entered as the Model ID.

See Reasoning on OpenAI and Azure OpenAI for the response shape, multi-turn replay, and the parameters that are dropped on this path.

Inference

After adding the models, you can perform inference using an OpenAI-compatible API via the Playground or by integrating with your own application.

Infer Model in Playground or Get Code Snippet to integrate in your application