A real-time endpoint, also known as an online endpoint in Azure Machine Learning, is designed for synchronous, low-latency inference. It is deployed to compute infrastructure that is provisioned and remains active ("always running") to serve prediction requests on demand. This makes it suitable for interactive applications where an immediate response is required. The underlying compute resources are always ready to process incoming requests without the startup delay associated with other deployment types.

A Batch endpoint is used for asynchronous, long-running inference on large datasets; its compute resources are typically provisioned only for the duration of the job.

Offline endpoint is not a standard term for a type of endpoint in Azure Machine Learning; inference can be done offline, but it's not an endpoint category.

Scheduled endpoint is not a distinct endpoint type; batch endpoint jobs can be scheduled, but the endpoint itself is not inherently "scheduled."

1. Microsoft. (2023). What are Azure Machine Learning endpoints?. Microsoft Learn. Retrieved from https://learn.microsoft.com/en-us/azure/machine-learning/concept-endpoints. (The document states, "Online endpoints are endpoints that are used for online (real-time) inferencing... They're always on and ready to receive new data to process.")

2. Microsoft. (2023). Deploy and score a machine learning model by using an online endpoint. Microsoft Learn. Retrieved from https://learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-online-endpoints. (This guide details the use of online endpoints for "real-time scoring.")