I just found out that multiple models of SSAS Tabular cannot be connected to one PBIX file. So my question is, should I deploy all the tables in one model or create several models where I will have some of the tables duplicated? Which will give me better performance?
You can create single big model which contains all what you need. Tables in model can be connected by some relation or not (in model you can have some independent part/group/island of tables without any connection), but I recommend use perspectives to avoid misunderstanding in large model.
From query performance there no difference if you have some tables in multiple models or in single model (but be carefully if all tables are connected by relations). Also there no difference about processing performance, every object you can process independently. And there are two pros: you can connect and use from single Power BI report, and you can save memory usage (and process cpu time) by duplicated tables in more models.
By my opinion, there is single cons, it is about development and maintenance, is more difficult to develop and keep developing cycle for single large model (especially if is developed by multiple developers) than more simple independent models.
To my knowledge, it would be better to create several models than one model. For each model, optimize it for its specific purpose. And it will result in a much better performance than trying to fit all requestments into one model.
Then you can create a report for each model and then create a dashboard based on them.
If this posthelps, then please considerAccept it as the solutionto help the other members find it more quickly.