You are in the process of deploying a new version of a model to a production Vertex AI endpoint that is actively serving user traffic. Your goal is to direct all user traffic to the new model while minimizing any disruption to your application. How should you proceed to achieve this objective?

Pick an option to reveal the answer.