Soot prediction in realistic systems is one of the most challenging problems in theoretical and applied combustion. Soot formation as a chemical process is very complicated and not fully understood up to the moment. The major difficulty stems from the chemical complexity of the soot formation processes as well as its strong coupling with the other thermochemical and fluid processes that occur simultaneously. Soot is a major byproduct of incomplete combustion, having a strong impact on the environment, as well as the combustion efficiency. Therefore, it needs to be predicted in realistic configurations in an accurate and yet computationally efficient way. In the current study, a new soot formation subgrid model is developed and reported here. The new model is designed to be used within the context of the Large Eddy Simulation (LES) framework, combined with Linear Eddy Mixing (LEM) as a subgrid combustion model. The final model can be applied equally to premixed and non-premixed flames over any required geometry and flow conditions in the free, the transition, and the continuum regimes. The soot dynamics is predicted using a Method of Moments approach with Lagrangian Interpolative Closure (MOMIC) for the fractional moments. Since, no prior knowledge of the particles distribution is required, the model is generally applicable. The effect of radiation is introduced as an optically thin model. As a validation the model is first applied to a non-premixed non-sooting flame, then a set of canonically premixed flames. Finally, the model is validated against a non-premixed jet sooting flame. Good results are predicted with reasonable accuracy.