Gilbert Quevauvilliers performs some testing:
In this blog post I am going to show you how I completed the automated testing and then the results where I am going to compare Direct Lake, Import and DirectQuery and which one appears to be the best.
As always, your testing may very or be different to my tests below.
I would highly recommend that you use the method I have used and apply this testing technique to your data to understand and validate which semantic model would be best for your capacity.
Click through for details on the tests, query durations, and how the three major modes of data loading into Microsoft Fabric semantic models (Import, Direct Lake, Direct Query) fare.