We're happy to announce the June 2025 release of Tabular Editor 3. We have packed this release with capabilities that help BI professionals build better semantic models faster in Microsoft Fabric, especially for those leveraging Direct Lake. So, let’s dig into how Tabular Editor now manages OneLake, composite models, and all the other improvements.
Watch the June 2025 release video on YouTube
This release leads with broader Direct Lake support.
We are excited to share that the experience of creating Direct Lake models has gotten a massive lift. Whether you are developing Direct Lake models from a Fabric Lakehouse or Warehouse, Tabular Editor now makes it a lot easier to find, set up, and use your Fabric tables and files. Moreover, we now support Direct Lake on OneLake through the AzureStorage.DataLake Power Query connector.
Direct Lake models require a Fabric Capacity and can only be created and managed in the Enterprise Edition of Tabular Editor 3.
The Import Table Wizard has been expanded with new data sources for Fabric Lakehouses and Warehouses to automatically let you connect to the objects you have access to in Fabric – you no longer need to search for, copy, and paste the SQL Endpoint to your Lakehouse or Datawarehouse!
After selecting the tables, the final page of the Import Table Wizard allows you to choose the partition mode that fits your needs; that is, do you want to use your Fabric Lakehouse or Warehouse tables in:
For a detailed presentation on choosing between DL/OL and Import mode, see this excellent (2-hour!) conference session by Marco Russo: Direct Lake vs Import vs Direct Lake+Import | Fabric semantic models (May 2025) - SQLBI
We've also written a short article with an overview of the various storage modes, the difference between Direct Lake on OneLake vs. Direct Lake on SQL, and what to be aware of.
With Tabular Editor 3.22.0, creating composite models that seamlessly blend Direct Lake and Import mode tables is as easy as choosing the right storage mode for your table. As shown above, Tabular Editor allows you to select which storage mode your table should use, but you do not have to stick to one mode per model. You can mix and match the partition modes, for example, Direct Lake for massive fact tables and Import for dimension tables.
Composite models are supported only when using Direct Lake over OneLake. Models using Direct Lake over SQL must first be converted to Direct Lake over OneLake. Here is a script that helps you perform this conversion.
In a composite model (Direct Lake + Import) a table can only have one storage mode. Direct Lake tables will always have one partition, whereas import mode can have multiple partitions.
For more information on setting up this new type of composite model with Tabular Editor 3, please read this article.
We've made several other improvements as well in the June 2025 release. Here are two highlights:
Thanks to the June 2025 Power BI Desktop update, external tools now have full write access to the semantic model hosted inside Power BI Desktop. As such, we have removed the restrictions Tabular Editor normally apply, when connected to Power BI Desktop as an external tool. For Desktop Edition users of Tabular Editor 3, this is a massive upgrade as it ensures that all semantic modeling capabilities are available in Tabular Editor 3, even previously inaccessible capabilities such as modifying M expressions, adding/removing tables, or setting up tables with multiple partitions. You can even copy/paste entire tables between models!
Power BI Desktop’s Power Query UI currently has a few limitations regarding Power Query expressions in table partitions. Specifically, tables with multiple partitions will disappear from the Power Query view. This will most likely be rectified in a future update. You can still view the partitions in the Model-, DAX- and TMDL views.
For more information on what this means, see this blog.
We have also made a subtle but helpful improvement to the DAX authoring experience.
You can now find measures, columns, and tables more easily in the auto-complete dropdown, even if you don't type out the exact words in the object name.
For example:
This word-based search makes it quicker to locate the right measure, even in large models with dozens of similarly named items.
The upgrade to the newest AMO/TOM library should fix a number of persistent authentication issues that some users have experienced when connecting to semantic models, often in B2B scenarios or when one or more Work & School Accounts are connected in Windows. This should mean a more reliable sign-on flow with fewer credential prompts. We want to thank Microsoft for prioritizing this fix in their library.
We have made multiple bug fixes in 3.22.0 and 3.22.1; please see our release notes for full details.
A frustrating issue that occasionally froze the Data Refresh view has been squashed. See this GitHub issue for details: #1461
Version 3.22.1 fixes an issue where the Data Refresh Pane would clear out and not show any refreshes.
Shortcuts now work even when the “What’s New” page is focused. And when assigning a shortcut already in use, you’ll get a warning, and the old mapping will be cleared to avoid conflicts.
See the complete list of updates in our release notes:
If you have any feedback or experience any bugs, please reach out in our community forum. We look forward to hearing from you.
That's the June 2025 release: a much smoother Direct Lake experience, composite Direct Lake + Import models, full modelling against Power BI Desktop, and a batch of DAX and authentication improvements. Grab 3.22.1, read the Direct Lake guidance, and let us know in the forum what you'd like next.