The correct option is Cloud Run because it is a fully managed and elastic service that runs container images behind an HTTPS endpoint, scales down to zero when idle, and can rapidly scale to handle short bursts such as about 4,000 requests per minute.

It executes stateless containers and provides automatic request based autoscaling with configurable concurrency so it can absorb promotional traffic spikes and then scale back to zero when requests stop. It provisions HTTPS by default and is well suited to expose a simple tool endpoint that an external generative AI agent can call securely.

Cloud Functions is serverless but it is designed for code written as functions rather than running an arbitrary container image, so it does not fit the requirement to deploy an existing container as is.

GKE Autopilot can run containers but it is a Kubernetes platform that introduces cluster concepts and does not natively scale workloads to zero by default, which adds operational overhead and misses the strict scale to zero requirement.

Vertex AI Model Garden is for discovering and consuming foundation models and it is not intended to host a custom containerized HTTPS API, so it does not match the use case.

Full AWS Practitioner Certification Question