Compressor corner stall is a phenomenon difficult to predict with numerical tools but essential to the design of axial compressors. Predictive methods are beneficial early in the design process to understand design and off-design limitations. Prior numerical work using Navier–Stokes computational methods has assessed the prediction capability for corner stall. Reynolds-averaged Navier–Stokes (RANS) simulations using several turbulence models have shown to over-predict the region of corner hub stall where large eddy simulations (LES) and detached eddy simulations (DES) approaches improved the airfoil surface and wake pressure loss prediction. A linear compressor cascade designed and tested at Ecole Centrale de Lyon provides a good benchmark for the evaluation of the accuracy of numerical methods for corner stall. This paper presents results obtained with Lattice-Boltzmann method (LBM) coupled with very large-eddy simulations (VLES) approach of PowerFLOW and compares them with the experimental and numerical work from Ecole Centrale de Lyon. The ability to achieve equivalent accuracy at a lower computational cost compared to LES scale resolving methods can enable multi-stage design and off-design compressor predictions. A methodical approach is taken by first accurately simulating the upstream flow conditions. Geometric trips are modeled upstream on the endwalls to match both the mean and fluctuating inflow boundary layer conditions. These conditions were then applied to the simulation of the linear compressor cascade. The benchline experimental study includes trips on both the pressure and suction of the airfoil. These trips are also included for the current simulation. The significance of capturing both inflow conditions and including trips on the airfoil is assessed. Detailed comparisons are then made to airfoil loading and downstream losses between experiment and previous RANS and LES simulations. LBM-VLES corner stall results of pitchwise averaged total pressure match LES agreement relative to experimental data at 50 times lower computational cost.