This study concerns optimization of shapes, locations, and dimensions of internal cooling passages within a turbine vane under severe environments. The basic aim is to achieve a design that minimizes the average temperature and ensures the structural strength. Considering the prohibitive computational cost of 3D models, numerical optimization process is performed based on 2D cross-sectional models with available experimental temperature data as boundary conditions of thermomechanical analysis. To model the cooling channels, three kinds of shape configurations, i.e., circle, superellipse, and near-surface holes, are taken into account and compared. Optimization results of 2D models are obtained by using a globally convergent method of moving asymptotes (GCMMA). Furthermore, full conjugate heat transfer (CHT) analyses are made to obtain temperature distributions of 3D models extruded from 2D ones by means of shear stress transport (SST) k-ω turbulence model. It is shown that optimization of cooling passages effectively improves the thermomechanical performances of turbine vanes in comparison with those of initial C3X vane. The maximum temperature of optimized vane could be reduced up to 50 K without degrading mechanical strength.