ProductsGatewayReliability

Fallback

Try alternate models in order after the current route exhausts its eligible attempts.

For the complete list of all request parameters, see API reference.

Respan tries the models in fallback_models in the order you specify after the current route exhausts its eligible attempts. Preflight rejections and failures marked fail-fast do not advance through the model list. See Retries and fallback order for the execution sequence and reliability matrix.

Fallback reduces the impact of a provider or model failure, but it does not guarantee availability. The gateway returns as soon as one model succeeds; if every eligible model fails, it returns the terminal error.

Go to Settings -> Fallback -> Click on Add fallback models -> Select the models you want to add as fallbacks.

You can drag and drop the models to reorder them. The order of the models in the list is the order in which they will be tried.

Fallback Page

How fallback interacts with retries

Retries are applied to the current route first. After its attempt budget is exhausted, the gateway advances to the next eligible fallback model. Each fallback has its own route attempts, so longer chains and higher retry counts increase worst-case latency.

Use Retries and fallback order to choose an attempt count and initial delay, and review Limits for requests that can be rejected before any provider call.