試験MLA-C01 トピック1 問題93 スレッド

Amazon SageMaker multi-model endpoints (MME) are designed to host multiple models behind a single endpoint, dynamically loading models into memory on demand. AWS documentation explicitly recommends MME as the most cost-effective solution when multiple models share the same ML framework and inference container.
With MME, SageMaker loads models from Amazon S3 only when they are invoked and unloads idle models automatically. This dramatically reduces the number of instances required and avoids paying for always-on resources for infrequently used models.
Multi-container endpoints are intended for inference pipelines or ensembles and require all containers to be loaded at startup, which increases cost. Deploying separate endpoints for each model results in the highest cost due to duplicated infrastructure.
AWS best practices clearly position multi-model endpoints as the optimal choice for reducing inference costs when hosting many models with similar runtime requirements.
Therefore, Option B is the correct and AWS-verified solution.