Zirconium-based metal–organic polyhedra (ZrMOPs) are attractive due to their high stabilities, low-cost building blocks, and solubilities relative to their metal–organic framework analogues (e.g., UiO-66); although these sorts of self-assembled cages often form single thermodynamic products, ZrMOP architectures are typically plagued by the formation of both V₄L₆ “tetrahedra” and V₂L₃ “lanterns” (V = vertex, L = ligand) as coproducts of their syntheses. In this work, we demonstrate a ligand-exchange strategy to isolate previously inaccessible phase-pure ZrMOPs using two different dicarboxylate donors. We also describe characterization methods that can be used to discriminate between the two architectures to confirm our approach provides synthetic selectivity. The phase-pure materials were found to have drastically different Brunauer–Emmett–Teller (BET) areas, with lanterns exhibiting significantly smaller surface areas (4–20 m²/g) than the tetrahedral architectures (393–605 m²/g), irrespective of counterions or bridging dicarboxylates. By obviating mixed-phase products of synthesis, our generalizable ligand-exchange pathway to phase-pure ZrMOPs enables systematic fundamental studies and will advance the functional use of these materials.