Simulations are performed for flow over circular cylinder at M = 0.2, Re = 5000 to assess the predictive capability of turbulence models and numerical schemes available in Loci-CHEM for the farfield acoustics predictions. A complimentary DNS is also performed for validation. Grid and time step study shows that a grid consisting of 7.5M cells and a non-dimensional time step size of 0.165 is sufficient for the hybrid RANS-LES simulations. Results demonstrate that numerical schemes and turbulence models play a critical role in the acoustics generation and propagation. Low-dissipation optimization gradient reconstruction (OGRE) scheme is found to be suitable for the acoustics predictions, but turbulence models show large differences. SST predicts only large-scale fluctuations. Dynamic hybrid RANS-LES (DHRL) performs better than hybrid RANS-LES (HRL) for the prediction of flow separation and turbulent coherent structures, but acoustics predictions are similar to URANS. HRL is found to be under predictive for acoustics predictions. MILES performs the best, suggesting that turbulence models are too dissipative. The future research will focus on implementation and assessment of wall-modeled LES models to improve the predictions.