About a year ago, we announced Power BI Premium support for large models in public preview. Today, we are excited to announce General Availability (GA), along with additional new features.
With large model support, a dataset in Power BI Premium is limited in size only by the Premium capacity that hosts the dataset. Excluding memory for query execution and processing, this typically means a dataset can grow up to 70-80% of the capacity’s total memory. Together with incremental refresh, this unlocks interactive analysis over very large datasets with minimal management overhead.
Large model support is one of the cornerstones to unblock customers looking for options to lower enterprise BI infrastructure complexity and total cost of ownership (TCO). Another is XMLA endpoints. Among other things, these critical capabilities open the door to consolidate tabular models from SQL Server Analysis Services and Azure Analysis Services on one common platform based on Power BI Premium. As already mentioned, this helps to reduce management overhead. Perhaps even more importantly, Power BI increases the discoverability of your enterprise BI solutions through such features as shared and certified datasets, which make it easy for you to promote trusted and authoritative datasets in your organization.
If you are planning to use XMLA -based tools to perform write operations, make sure you enable the large dataset storage format even for datasets that you wouldn’t necessarily characterize as large datasets because this storage format helps to boost XMLA write performance. You can change the storage format for a dataset in the dataset settings when you expand the Large dataset storage format section. It is also a good idea to make the large dataset storage format the default format for the workspaces where you store your datasets.
| Dataset settings | Workspace settings |
 |
 |
As part of the GA release, we are also introducing the following new capabilities:
- User interface enhancements. Users can conveniently convert existing datasets to the large storage format and set workspace defaults in the Power BI portal, as the above screenshots illustrate.
- Multi-Geo support for large models. Multi-Geo is a Premium feature that helps multinational customers address regional, industry-specific, or organizational data residency requirements. You can now deploy large models in regions other than the home region of your Power BI tenant.
- Bring your own encryption keys (BYOK) for large models. By default, Power BI uses Microsoft-managed keys to encrypt your datasets at rest, but you can also use your own keys in Power BI Premium. With BYOK for large models, you can control and revoke your organization’s keys to increase security and meet compliance requirements. For example, if you revoke the keys, your datasets are unreadable to the service within 30 minutes.
Please see https://aka.ms/LargeModels for further information such as availability by region, and how to set up your Power BI Premium capacity for large models.