Which mode should be used to implement the Research division’s semantic models?

The correct choice for implementing the Research division’s semantic models is Direct Lake. This mode is particularly effective for scenarios where large datasets need to be accessed and analyzed in real-time without the necessity of moving the data into a traditional data warehouse.

Direct Lake allows users to connect directly to the data lake, enabling efficient querying of structured and semi-structured data. This approach facilitates rapid access to the latest data directly from the source, maintaining the semantic layer's integrity while ensuring that users can perform analysis with a comprehensive dataset that's constantly updated. This is crucial in research environments where timely access to data is essential for decision-making and data-driven insights.

Other options may involve data being imported into a centralized system or warehouse, but such methods could lead to delays in data updates and require additional storage management. Methods like Live Connection and Direct Query provide some level of real-time access but may not offer the same efficiency and capabilities as Direct Lake when it comes to working with large volumes of diverse datasets typical in research applications.