The enumeration of finite models of first order logic formulas is an indispensable tool in computational algebra. The task is hindered by the existence of isomorphic models, which are of no use to mathematicians and therefore are typically filtered out a posteriori. This paper proposes a divide-and-conquer approach to speed up and parallelize this process. We design a series of invariant properties that enable us to partition existing models into mutually non-isomorphic blocks, which are then tackled separately. The presented approach is integrated into the popular tool Mace4, where it shows tremendous speed-ups for a variety of algebraic structures.