You are deploying a new version of a model to a production Vertex Al endpoint that is serving traffic. You plan to direct all user traffic to the new model. You need to deploy the model with minimal d

Creating an entirely new endpoint and updating applications would cause disruption and downtime during the switch, which goes against the requirement of minimal disruption.

Creating an entirely new endpoint and updating applications would cause disruption and downtime during the switch, which goes against the requirement of minimal disruption.

The optimal approach to deploy a new model version with minimal disruption is to upload the new model to the same model resource, then deploy this new model alongside the existing one on the endpoint and manage traffic splitting to ensure a smooth transition without downtime.

Setting a new model as the 'default version' is not the standard Vertex AI Model Registry concept for managing live endpoint traffic splitting; managing versions and traffic occurs at the endpoint deployment level.