Exam AWS Certified Machine Learning Engineer - Associate MLA-C01 topic 1 question 103 discussion

The correct solution for deploying the ML model on Amazon SageMaker with payload sizes between 100 MB to 300 MB and processing requirements of 60 minutes or less is to use Asynchronous inference (option B). This is the most suitable choice because asynchronous inference is specifically designed to handle large payload sizes (up to 1 GB) and long processing times (up to 15 minutes per request), while real-time inference has payload limitations of 6 MB, serverless inference has a 4 MB limit, and batch transform is better suited for offline processing of entire datasets rather than individual requests.

Agree with B. In general, real-time inference supports payloads up to 5 MB for synchronous requests, while asynchronous inference can support larger payloads, often up to 5 GB. The use case in this questions involves inference payloads of 100 MB to 300 MB and needs to be processed in under 60 minutes, Asynchronous Inference is the best choice for handling large payloads without strict real-time requirements.