Thanks for the answers. Really appreciated! I am analyst and I deal with databases, but in a very specific way. The way I tried to explain to myself the difference is that OLAP is used if you have a lot of fine-grained hierarchies accessed by a big number of people. In this scenario it would make sense to make all those aggregations ahead of time because a columnar store would have to calculate them each time somebody slices and that would be waay to expensive, even if you have state of the art caching.
Power BI Desktop (and hosted data models on the PowerBI.com cloud service) is also powered by the Tabular engine and utilizes DAX as a query and modeling language.
Microsoft is pushing Tabular hard. It's probably the right choice for most organizations, and at worst serves as a great prototyping tool for a Multidimensional model (when we have clients who need Multidimensional for whatever reason, we often still do initial development and pilots on Tabular).