A three-dimensional, full-scale, transient computational fluid dynamics (CFD) model was developed to simulate the fluid flow, phase interfaces and slag-eye characteristics as well as the slag-metal reactions and desulfurization efficiency in the gas-stirred ladles. The volume of fluid (VOF) model, discrete phase model (DPM), and realizable k-ε model were applied to describe the argon/steel/slag/air multiphase evolution and the turbulent flow. The model was validated by comparing the simulation results with the experimental data based on a water model and an industrial gas-stirred ladle. The multiphase VOF model and the DPM-VOF model were applied to simulate the multiphase flow in the water model and gas-stirred ladle, respectively. The comparisons of the simulated and experimental results demonstrated that the DPM-VOF coupled model is more accurate for predicting the gas-liquid multiphase flow phenomena. The effects of the gas flow rate, bubble characteristics, slag layer thickness, slag viscosity and plug configurations on the flow characteristics and mixing phenomena in the ladles were also investigated. A CFD-reaction kinetics fully coupled model was developed and validated to predict the slag-metal reactions and desulfurization behavior in an industrial gas-stirred ladle. A quick modeling approach was developed by uncoupling the slag-metal reaction kinetics computations from the CFD simulation. It was shown that the computational time of this uncoupled predictive approach was decreased by at least 100 times for each case study in comparison with the CFD-reaction kinetics fully coupled model. The validated model was also applied to investigate the effects of the steel and slag compositions on the slag-metal reactions and on the desulfurization efficiency. Finally, a CFD modeling approach capable of predicting the transport and removal behavior of inclusions in a gas-stirred ladle was developed with considering a fluctuant top slag layer. The inclusions were tracked by DPM model, and the movement of individual inclusion was traced through computing their particle trajectories. The effects of the inclusion size, gas flow rates, and injected bubble diameters as well as various removal mechanisms including slag capture, bubble attachment, and ladle wall adhesion on the removal of inclusions were also investigated.