Which service allows users to create custom machine learning models on AWS?

Amazon SageMaker is the service that enables users to create custom machine learning models on AWS. It provides a fully integrated development environment for building, training, and deploying machine learning models at scale. SageMaker simplifies the process by offering built-in algorithms, automated model tuning, and an extensive set of tools for data labeling, data preprocessing, and model evaluation. This allows data scientists and developers to focus on the model creation and training process rather than the underlying infrastructure, making it easier to develop high-quality machine learning solutions.

The other services listed serve different purposes. AWS Glue is primarily an ETL (Extract, Transform, Load) service used for preparing data for analytics. Amazon Comprehend is a natural language processing service that helps analyze text and extract insights from it, such as sentiment and entity recognition, but it does not facilitate the creation of custom machine learning models. Amazon Rekognition focuses on image and video analysis, providing features like object detection, but is not designed for creating custom ML models tailored to specific use cases. Thus, SageMaker stands out as the dedicated machine learning model development service on AWS.