Current vehicle aerodynamic development makes extensive use of Computational Fluid Dynamics (CFD) to enable cost-effective design and parametric exploration. Although larger-scale, high fidelity simulations are increasingly popular, practical Reynolds number ranges (105-108) necessitate hybrid modelling approaches which offer alternatives to fully-resolved Large-Eddy Simulation (LES) for predictive capability of separated, turbulent flows.A detailed aerodynamic investigation is conducted over a SAE Notchback model which experiences subtle pressure-induced separation near the roof-backlight junction and shear layer roll-up and a low-pressure wake at the rear. Computational results were generated by a novel Detached-Eddy Simulation (DES) algorithm implemented in a high-order, compressible CFD code FLAMENCO.The study presents a novel hybrid turbulence modelling algorithm which combines a high-order, enhanced Spalart-Allmaras model with an Implicit LES, to effectively relax associated grid resolution requirements and eliminate the need for explicit subgrid scale models. This varies the effective length scales within the governing transport equations and the modelled eddy viscosity, with minimal user-interaction or zonal definitions.Two variants of the notchback are tested to evaluate the impact of including an underbody diffuser. Simulations captured majority of the mean and unsteady behaviour of the flow even on relatively coarse meshes (2-8 million cells) for automotive CFD.