# See which model answered from the AI Gateway response headers

> Every AI Gateway response now carries its request ID, its attempt count, and the model that served it.

- Date: 2026-09-02
- Shipped by: mohamed
- Products: AI Gateway
- URL: https://ngrok.com/changelog/ai-gateway-response-headers

When the gateway fails over, the provider's response body can name a different model than the one you asked for. Three response headers now tell you what happened:

- `Ngrok-AIG-Request-Id` is the request's ID, on every response, including rejections and errors.
- `Ngrok-AIG-Attempts` counts the upstream attempts the gateway made.
- `Ngrok-AIG-Model` names the upstream model that produced the response.

Browser clients can read all three through CORS. They're ngrok's headers, so you get the same values with ngrok credits or your own provider keys. The native Gemini `generateContent` and `streamGenerateContent` endpoints return them too.

## Get started

1. Send any request through `https://gateway.ngrok.ai`.
2. Read the headers on the response. In the OpenAI SDK, `.withResponse()` returns them next to the data.
3. Treat `Ngrok-AIG-Model` as the answer to "who replied?" See [response headers](https://ngrok.com/docs/ai-gateway/how-it-works#response-headers).
