If you're talking about other 3D objects (such as can be found on the web) then the answer is also resounding yes for Version 2, IF you have some knowledge about what you're doing and you can get or export the model in VRML Format.
These VRML objects become native FrameForge 3D Studio 2 objects simply by placing them in your FrameForge object library and running FrameForge. If you used specific struturing and naming conventions, you can even end up with an object that has hinges and user-swappable parts just like any other normal FrameForge Object.
However, before you rush off and try to bring in a billion new objects, please read the rest of this message.
Our most complex objects (such as actors and picture cars) have about 15-17K polygons, most simple objects should have no more than 1-2K polygons and the fewer the better in terms of rendering. It's not uncommon, however, for 3D objects which you can find on the web to have 50-60K polygons for even something "simple" like a coffee pot or plant. Simply put, most 3D objects found on the web are often not optimized for real-time rendering so bringing them into a set can slow down the program significantly.
Secondly, FrameForge uses what is called backface culling to speed up the display. This means that it does not draw whatever is designated as the "back" side of a given surface. Again, most 3D objects you'll find on the web were not designed with backface culling in mind and if you bring them in you'll find weird gaps in their image as shown in this image of a battleship. (And the original file was so badly designed, I had to clean it up to even show you something intelligible!)
The function that describes which side of a part is "front" and which is "back" is called NORMALS and while vitually all 3D Modelling software the ability to adjust an object's normals, it can be a time consuming process.
Full details on this can be found in the program manual on pages 150-154