The computational modelling of soot formation and destruction during the combustion process is one of the most challenging topics in combustion research. This paper reviews the numerical soot models constructed at different length scales, including macroscale, mesoscale, and microscale. The four key stages of soot evolution, including nucleation, surface growth and coagulation, agglomeration, and oxidation, are first described with the generally accepted mathematical formulations in each stage explained. Different computational frameworks and their pros and cons are then reviewed, including the one-equation empirical soot model (macroscale), two-equation semi-empirical soot model (macroscale), different variations of population balance model (mesoscale), discrete element model (microscale), and molecular dynamics model (microscale).It is concluded that the accuracy required and the computational cost available are the two major influencing factors to be considered when selecting the appropriate computational model. The user needs to assess the priorities in their specific application and evaluate different modelling options to find the optimal balance between the level of accuracy and computation resources required.