7. Make your first API call

The gateway exposes an OpenAI-compatible chat completions endpoint on your organization's path. If you've used the OpenAI API, this will feel familiar — you only change the base URL and the token.

Endpoint

POST https://ai-gateway.optum.com/{org}/v1/chat/completions

Replace {org} with your organization slug (from step 1, e.g. acme-health).
Send Authorization: Bearer <access_token> from step 6.
The request/response bodies follow the OpenAI chat completions schema.
The model value is the provider-prefixed id from your model list (and GET /{org}/v1/models) — e.g. azure/gpt-4.1-mini. The prefix selects the upstream provider, so copy the exact id the console shows.

curl

curl https://ai-gateway.optum.com/acme-health/v1/chat/completions \
  -H "authorization: Bearer $TOKEN" \
  -H "content-type: application/json" \
  -d '{
    "model": "azure/gpt-4.1-mini",
    "messages": [{ "role": "user", "content": "Say hello in one word." }],
    "max_tokens": 20
  }'

A successful response:

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "gpt-4.1-mini-2025-04-14",
  "choices": [
    { "index": 0, "finish_reason": "stop",
      "message": { "role": "assistant", "content": "Hello" } }
  ],
  "usage": { "prompt_tokens": 12, "completion_tokens": 1, "total_tokens": 13 }
}

Streaming

Set "stream": true to receive the response as Server-Sent Events (data: chunks terminated by data: [DONE]):

curl https://ai-gateway.optum.com/acme-health/v1/chat/completions \
  -H "authorization: Bearer $TOKEN" \
  -H "content-type: application/json" \
  -d '{
    "model": "azure/gpt-4.1-mini",
    "messages": [{ "role": "user", "content": "Count to five." }],
    "stream": true
  }'

OpenAI Python SDK

Point the SDK's base_url at your organization path and pass the token as the API key:

from openai import OpenAI

client = OpenAI(
    base_url="https://ai-gateway.optum.com/acme-health/v1",
    api_key=TOKEN,  # the bearer access_token from step 6
)

resp = client.chat.completions.create(
    model="azure/gpt-4.1-mini",
    messages=[{"role": "user", "content": "Say hello in one word."}],
)
print(resp.choices[0].message.content)

Common errors

Status	Meaning	Fix
`401 Unauthorized`	No token, or an expired / malformed token.	Send `Authorization: Bearer <token>`; refresh if expired (step 6).
`403 Forbidden`	The token is valid but its client isn't authorized for this organization.	Confirm you're calling your own org's `{org}` path and that the credential belongs to that org (step 5).
`500` "failed to get balancer instance"	The requested `model` isn't enabled on this organization's route.	Enable the model and Publish (step 4), or send a model that's enabled.
`404 Not found` (HTML)	The path or org slug is wrong.	Check the base URL and `{org}` slug.

That's the full path — you now have a working, authenticated chat completion through the AI Gateway.