Honestly it largely depends on the scene. Sometimes you want to trade vertex / poly count for object number, sometimes the other way around, depending on what are currently bound by. Sometimes it’s draw calls, sometimes it’s fillrate and so on. Sometimes the sole fact of having 1 more mesh is what hurts you most, sometimes it’s the expensive pixel shader that’s on it. That kind of question rarely can be considered ‘answered and settled for good’.
You also have to take the account of how occlusion works - it would work faster with just 1 mesh instead of 5. If it’s just 1 on the map you won’t notice it, but when you go into hundreds it might be noticeable. Dynamic lighting and shadows would also suffer with more objects to calculate. Depends on how you want to use it - again back to the scene.
Besides, you haven’t mentioned how many polys the right example has.
Thanks man, your answer helped me double my FPS! I went with just 1 single model