The model stays the same.
Infera serves only the exact model you request. It never substitutes another model.
Exact model
Availability in the catalog indicates whether a model can receive requests now.
Availability
A model can receive traffic only while it is active, verified, healthy, priced and the inference service is ready. The catalog provides a safe reason when a model is pending or unavailable.
Requests and failover
If a request cannot be completed, Infera may retry the same model before any output is sent. After output begins, the response continues without switching to a different model.
If no eligible route is available, the request fails with model_unavailable and nothing is charged.