Reference · API

Azure OpenAI chat completions

Last updated Jul 17, 2026

Run **OpenAI-compatible chat** on Microsoft Azure — enterprise SSO, private networking, and regional deployment.

#When to use

Org policy requires Azure instead of public OpenAI

Private endpoints and Azure AD auth for internal apps

Predictable regional data residency

Same message shape as OpenAI chat completions; URL and auth differ.

#Endpoint (pattern)

POST https://{resource}.openai.azure.com/openai/deployments/{deployment}/chat/completions?api-version=2024-02-15-preview

`deployment` is your Azure **deployment name**, not always the raw model name.

#Headers

api-key: $AZURE_OPENAI_API_KEY
Content-Type: application/json

Or Azure AD bearer token when using managed identity.

#Request shape (sketch)

{
  "messages": [
    {"role": "system", "content": "You are concise."},
    {"role": "user", "content": "Explain vectors simply."}
  ],
  "temperature": 0.2,
  "max_tokens": 512,
  "stream": false
}

#Python sketch

from openai import AzureOpenAI
client = AzureOpenAI(
    azure_endpoint="https://YOUR_RESOURCE.openai.azure.com",
    api_key="...",
    api_version="2024-02-15-preview",
)
resp = client.chat.completions.create(
    model="YOUR_DEPLOYMENT_NAME",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

#Common pitfalls

| Pitfall | Fix |

|---------|-----|

| Using public OpenAI base URL | Point SDK at `azure_endpoint` + deployment |

| Confusing deployment vs model name | Create deployment in Azure portal; use that ID |

| Stale `api-version` | Pin version in config; test on upgrade |

| Streaming without SSE proxy | Same server-side streaming rules as OpenAI |

Learn by doing → Try the interactive lesson →

← All API docs