<p>Tow-steered composites offer the possibility to tailor and enhance the mechanical performance of lightweight structures thanks to their larger design space compared to straight-fiber composites. This work proposes a scalable low- to high-fidelity methodology to retrieve the fiber orientations that optimize strain and stress distributions in variable stiffness plates. An optimization algorithm that combines global and local search strategies solves unconstrained and manufacturing-constrained problems. The structural models are generated through the Carrera Unified Formulation, which permits tuning the accuracy of the solution by selecting the order of the structural theory employed. The results show differences in the optimal stacking sequences as free-edge effects, local distortions, and 3D stress states are involved in the objective functions. Additionally, differences in the prediction of the quantities of interest are found between low-to-refined equivalent-single-layer—including the particular cases of the classical plate theory and the first-shear order deformation theory—and high-fidelity layer-wise models.</p>