Moving Semantic Model

Moving a large semantic model (50 GB) from a Dev to a Prod workspace in Power BI without using a PBIX file is necessary here due to dataset size. When working with models this large (typically Direct Lake or Import mode on Fabric), PBIX import/export is not viable due to file size limitations and platform restrictions.

Option 1: Deployment Pipelines

• You must have Power BI Premium or Fabric Capacity.
• Both workspaces must be assigned to a capacity.
• Your dataset must be deployed using Power BI datasets, not AS models or third-party tools.

 Steps:

1. In Power BI Service, create a Deployment Pipeline (Dev > Test > Prod).
2. Add your Dev workspace to the pipeline as the Development stage.
3. Assign the Prod workspace to the Production stage.
4. Run a deployment from Dev to Prod.
• This will copy dataset + data (unlike PBIX export).
• You can configure rules to parameterize things like data sources.
5. Confirm the deployed dataset appears and is functional in the Prod workspace.

Option 2: XMLA Endpoint – Backup and Restore via SSMS

Only valid for Fabric/Premium workspaces with XMLA Read/Write enabled.

• Premium/Fabric Capacity workspace
• XMLA endpoints must be read/write enabled
• You need SQL Server Management Studio (SSMS) installed.

Steps:

1. Connect to powerbi://api.powerbi.com/v1.0/myorg/{workspaceName} in SSMS.
2. Locate the model in the Dev workspace, right-click > Back Up to a .abf file.
3. Connect to the Prod workspace via SSMS.
4. Right-click > Restore > Provide the .abf file and specify a new or existing name.

Please mark this post as a solution if it helps you. Appreciate Kudos.

Hi Akshatraj_Adig 

When you're working with a large semantic model—like your 50 GB dataset—in Power BI, you can't move it using a .pbix file because of size limitations and because large models are typically created and managed directly in the Power BI Service using deployment pipelines or APIs. To move a semantic model with its data from a development workspace to a production workspace, the best practice is to use Power BI Deployment Pipelines, which are designed specifically for promoting content (including semantic models) across dev, test, and prod stages without needing to republish or reprocess data.

However, for this to work, both your dev and prod workspaces must be assigned to Power BI Premium capacity (or Fabric capacity), and the semantic model must be stored in a workspace connected to a deployment pipeline. When you deploy through the pipeline, it can reuse the existing model and partitions, avoiding a full data reload. If you're not using deployment pipelines or need more control (e.g., for custom CI/CD scenarios), you can also automate the process using Power BI REST APIs—specifically the Export and Import APIs for large models—which allow you to export the semantic model from one workspace and import it into another along with its data.

If you’re only using the PBIP structure or don't yet have deployment pipelines set up, you’ll likely hit a roadblock, as PBIP and .pbix deployments don’t carry over data—they only move metadata. So, the solution is to either enable deployment pipelines with Premium capacity or use REST API-based automation to move the model and data efficiently from dev to prod.

Hi Akshatraj_Adig ,

Thank you Poojara_D12, andrewsommer  for the helpful responses. I completely agree with the points shared. Since the semantic model is 50 GB, exporting via PBIX isn’t feasible. The two supported approaches are  Deployment Pipelines, which is recommended and most seamless if both workspaces are on Premium or Fabric capacity, and XMLA based backup and restore using SSMS for more advanced control. Both methods support moving the model along with its data.

I've also attached the official Microsoft documentation below for better understanding
Deployment Pipelines Overview
Overview of Fabric deployment pipelines - Microsoft Fabric | Microsoft Learn

If any of the responses above helped resolve your issue, please consider marking it as the accepted solution to make it easier for others facing a similar problem in the community.

Regards,

Yugandhar.