Recent advances in modeling primary atomization in order to enable accurate practical-scale jet spray simulation are reviewed. Since the Eulerian–Lagrangian method is most widely used in academic studies and industrial applications, in which the continuous gas phase is treated in the Eulerian manner and droplets are calculated as Lagrangian point particles, the main focus is placed on improvement within this framework, especially focusing on primary atomization where modeling is the weakest. First, limitations of the conventional methods are described and then novel modeling proposals intended to tackle these issues are covered. These new modeling proposals include the Eulerian surface density approach, and the hybrid Eulerian surface/Lagrangian subgrid droplet generation approach. Compared to conventional simple yet sometimes non-physical models, recent models try to include more physical findings in primary atomization which have been obtained through experiments or direct numerical simulation (DNS). Model accuracy ranges from one that still needs some adjustment using experimental or DNS data to one which is totally self-closed so that no parameter tuning is necessary. These models have the potential to overcome the long-recognized bottleneck in primary atomization modeling and thus to improve the accuracy of whole spray simulation, and may greatly help to improve the spray design for higher combustion efficiency.