Forum Discussion
Semantic Model Size in Pro Workspace Randomly Jumps From 300 MB to 800 MB – No Changes
I’m on a Power BI Pro license, working in a Pro workspace with a mixed model — Import mode plus a live connection to another semantic model.
When I first publish, the dataset size in the service is around 300 MB. A few days later—without making any changes or refreshing—it suddenly shows as 800 MB in the service.
If I export the model from the service and immediately re-publish it under a different name (no changes at all), the size goes right back to 300 MB. Auto Date/Time is turned off.
Has anyone else experienced this behavior or know what could be causing the size jump? Is the size actually jumping or is it just a bug?
Hi Lightsong90
What you’re seeing is a known quirk with composite or mixed models in Power BI Service, especially when combining Import mode with a live connection to another semantic model. The dataset size shown in the Service isn’t always a true reflection of the underlying storage footprint; sometimes the Service overestimates size because of how it tracks temporary objects, cached query results, or internal indexing after the model has been interacted with for a while. That explains why the dataset shows as ~300 MB right after publishing, then later appears to “inflate” to 800 MB even without refreshes or changes. When you re-publish the same PBIX, the Service effectively clears out those cached elements and recalculates the size correctly, dropping back down. In other words, the model itself isn’t really growing—this is more of a reporting/metadata bug in the Service than actual data bloat. Microsoft has had similar issues reported for composite models, so unless you see performance or refresh problems, you can treat the jump as cosmetic rather than a real storage issue.
6 Replies
Hi Lightsong90
What you’re seeing is a known quirk with composite or mixed models in Power BI Service, especially when combining Import mode with a live connection to another semantic model. The dataset size shown in the Service isn’t always a true reflection of the underlying storage footprint; sometimes the Service overestimates size because of how it tracks temporary objects, cached query results, or internal indexing after the model has been interacted with for a while. That explains why the dataset shows as ~300 MB right after publishing, then later appears to “inflate” to 800 MB even without refreshes or changes. When you re-publish the same PBIX, the Service effectively clears out those cached elements and recalculates the size correctly, dropping back down. In other words, the model itself isn’t really growing—this is more of a reporting/metadata bug in the Service than actual data bloat. Microsoft has had similar issues reported for composite models, so unless you see performance or refresh problems, you can treat the jump as cosmetic rather than a real storage issue.
Hi Lightsong90
This is AI generated and unverified, but I hope it helps:
Possible Causes
1. Service-Side Caching or Metadata Bloat
Power BI Service may be caching query results or metadata related to the live connection. This can inflate the reported size even though the actual imported data hasn’t changed.
- Why it resets on re-publish: Re-publishing clears the cached metadata and resets the model footprint.
- Why it grows over time: Usage patterns, query logs, or internal optimizations might be stored temporarily.
2. Hybrid Model Behavior
Mixed models (Import + Live) are relatively new and still evolving. The live connection might be causing the service to store additional metadata or intermediate results.
- Live connection to semantic model: If that model changes or gets queried heavily, Power BI might store more info in your dataset to optimize performance.
3. Model Snapshot or Versioning
There’s a possibility that Power BI is internally versioning or snapshotting parts of the model for reliability or rollback purposes.
- Not visible to users but could affect reported size.
4. Bug in Size Reporting
It’s also possible that the dataset size shown in the service is not accurate and reflects temporary or inflated values due to a bug.
- Especially if the re-published version is identical and shows the original size.
✅ What You Can Try
- Use XMLA endpoint to inspect the model metadata and see if anything unexpected is being stored.
- Monitor usage metrics to see if certain queries or users are triggering large cache footprints.
- Open a support ticket with Microsoft if this persists — especially since it’s reproducible and affects storage quotas.
Hi Lightsong90
While I understand it is a potential concern that the size of the semantic model grows, is there anything in particular that you're worried about with the size growth or their performance issues, etc? Or is it more just around the size of the model increasing? And if that is the case, then as long as it is under the 1GB size I would not be too concerned.
Hi Lightsong90
Thank you for reaching out to the Microsoft Fabric Forum Community.
GilbertQ wardy912 Thanks for the inputs. I hope the information provided by users was helpful. please try those, If you still have questions, please don't hesitate to reach out to the community.
Hi Lightsong90
I wanted to check if you had the opportunity to review the information provided by users. Please feel free to contact us if you have any further questions.
Hi Lightsong90
Hope everything’s going smoothly on your end. I wanted to check if the issue got sorted. if you have any other issues please reach community.
