Abstract: This study ‌evaluates‌ the predictive performance of ‌the‌ widely used Spalart-Allmaras (SA) and Shear Stress Transport (SST) turbulence models, ‌along with‌ the helicity-modified SA-Helicity and SST-Helicity models proposed by the research team, ‌for‌ complex compressor flows ‌using‌ the domestically developed multi-field coupled simulation platform AeroEngine Numerical Simulation (AENS). Two representative test cases, namely the highly loaded NACA65 K48 compressor cascade and the transonic axial compressor rotor NASA Rotor37, were employed to systematically assess the capabilities of these four turbulence models in capturing three-dimensional corner separation and other intricate flow phenomena. Results demonstrate that, for the NACA65 K48 cascade, the SA-Helicity and SST-Helicity models significantly improve the accuracy of total pressure loss prediction within corner separation regions, reducing the errors by 22.53% and 27.19%, respectively, compared to the original SA and SST models. For NASA Rotor37, the‌ SA-Helicity and SST-Helicity models ‌demonstrated‌ close agreement with experimental measurements in both overall performance and spanwise aerodynamic parameter distributions, ‌with‌ total pressure ratio errors near stall ‌decreasing‌ by 79.20% and 76.33%‌ relative to the SA and SST models. Overall, the SA-Helicity and SST-Helicity models developed on the AENS platform exhibit superior predictive capability for compressor aerodynamic performance and corner-separation topology, highlighting their potential for practical application in compressor aerodynamic design.