A model on Nanonets is nothing but a narrow AI model trained to process a particular document type. You can't process passports on the invoice model and vice versa. Similar documents can be processed using the same documents but we the highest accuracy when users use 1 model for 1 document type.
To further clarify, a document type is different from a variant or different template of a document. For example, you can process n number of different sellers on 1 invoice model if they are similar i.e they broadly use the same language/taxonomy/nomenclature.