Depends on the perspective. The predicate will always be an IRI. The object will either be an IRI or a literal, and all literals in RDF (as of RDF 1.1) are typed , though serialization formats like Turtle work with implied types.
There is also the option of blank nodes for objects, though in almost all implementations they are stand-ins for anonymous IRIs, so in some sense or another almost anything has "a" type.
If you have RDFS inference turned on, notably, just using ?p in a triple will imply
?p a rdf:Property .
In plain RDF you can say a property is of some other ?type but I think you can get in trouble with that if you want to do OWL DL inference and you might want to say something like
if baz is meant to be relationship then I was low key wrong in my comment (I thought baz was the kind of bar, which can be untyped). But I guess the relationship must be a property at least
There is also the option of blank nodes for objects, though in almost all implementations they are stand-ins for anonymous IRIs, so in some sense or another almost anything has "a" type.