Here I address the question of what unit of analysis of models philosophers should investigate. I argue that the philosophical investigation of models should be focussed on clusters or research questions, rather than on single models and their components, as has generally been done. I suggest that two specific philosophical questions, which attempt to solve the mystery posed by models, may have guided the interest of philosophers towards individual models and model components. One of the questions is the kind of representational relation that exists between models and their targets. The other is the role of models within the debate on realism. The modelling practice, as well as philosophical arguments that maintain that our models are incapable of fulfilling at once all the purposes we might have for them, are, nevertheless, compelling reasons to explore how models among themselves are related. Using models in the New Economic Geography (NEG) as example, I argue that models are used such that they establish what I call vertical and horizontal complementarities with other models. These are important to determine the epistemic import of models.