Question

Your team uses an Azure AI Foundry project and wants to deploy a Meta Llama model as a pay-as-you-go, per-token endpoint without provisioning or managing any GPU compute. You need to automate this deployment using the Azure CLI in a CI pipeline. Which action correctly provisions the endpoint?

Accepted Answer

Create a serverless API deployment for the model in the Foundry project, which exposes a token-billed endpoint requiring no managed compute

Answer

Create a managed online endpoint backed by a GPU compute instance and deploy the model to it

Answer

Purchase provisioned throughput units (PTUs) and reserve dedicated capacity for the model before deploying

Answer

Attach an Azure Kubernetes Service cluster to the project and deploy the model as a Kubernetes inference workload