SmartRetail stores dozens of inference models in an S3 model store and uses a SageMaker multi-model endpoint on an ml.m5.xlarge instance to host them. A developer needs to invoke a specific model named retail-v3-2025.tar.gz from the multi-model endpoint using the SageMaker runtime API. Which invocation approach will load and run that specific model at request time?

The correct approach for invoking a specific model on a SageMaker multi-model endpoint is to use the TargetModel parameter in the InvokeEndpoint API call or by setting the X-Amzn-SageMaker-Target-Model HTTP header. This explicitly tells the multi-model container which model artifact, identified by its relative path (e.g., retail-v3-2025.tar.gz), to load and use for the inference request. The multi-model container then dynamically loads this model from the designated S3 model store. Incorrect options: ModelName is not a valid parameter for specifying a model within a multi-model endpoint invocation. The Content-Type header specifies the format of the request body, not the target model. Creating a new single-model endpoint defeats the purpose of a multi-model endpoint, which is designed to host multiple models on a single instance for cost efficiency and simplified management. Multi-model endpoints are specifically designed to be targeted to specific models at request time.