The correct answer is Vertex AI.

Vertex AI provides managed online prediction endpoints that let you deploy models as low latency APIs for instantaneous predictions. It handles autoscaling, model versioning, traffic splitting, monitoring, and built in explainability so teams do not need to manage serving infrastructure.

Vertex AI integrates with AutoML and custom model containers and offers a purpose built experience for machine learning serving which matches the requirement for serverless scalable APIs for model inference.

Cloud Run is a serverless container platform that can host model containers however it does not provide the ML serving features that Vertex AI offers such as built in model monitoring, versioning, and explainability, so it is not the best managed ML serving choice.

Google Kubernetes Engine gives full control over clusters and workloads but it requires more operational management and it is not a serverless managed ML service, so it does not match the question intent.

Cloud Functions is an event driven serverless compute option with execution time limits and cold start behavior and it lacks the model serving features of Vertex AI so it is not suitable for high performance online model APIs.

Full AWS Practitioner Certification Question