AWS Certified AI Practitioner AIF-C01 Practice Question

Batch inference collects a sizeable dataset, runs predictions on that set in one job, and writes the results to storage. Because it is not answering interactive user requests, it can accept higher latency and does not need a continuously running endpoint. Real-time inference, by contrast, keeps an endpoint active to return single-record predictions within milliseconds. The distractors describe properties of real-time inference, misuse the term inference for model evaluation, or falsely claim that batch inference requires persistent endpoints.