Numerical model as a simulation tool was used to describe the pollutants transformation and degradation process in constructed wetlands (CWs). It can help provide insight into the “black box” and increase the understanding of the complex processes in CWs. In the last few decades, several process-based numerical models were developed to depict the pollutants removal processes in CWs, which include biochemical model, hydraulic model, reactive-transport model, plants model, clogging model, and coupling model combining two or more sub-models. However, there was a long way to go before fully understanding the decontamination mechanisms of CWs. On the one hand, single or a composite model coupling a small number of sub-models cannot fully reveal the decontamination processes. On the other hand, a comprehensive model including all sub-models of current cognition involves numerous parameters, most of which are interaction and cannot quantitatively determined, thus making the model complex and leading to diffuse interaction. Therefore, in order to describe the reaction processes in CWs more accurately, it is expected that all parameters should be quantified as far as possible in the future model. This study aims to provide a review of the numerical models of CWs and to reveal mechanism of decontamination. Based on the advantages and disadvantages of existing models, the study presented the improvement method and future research direction: (1) new detection/monitoring technique or computing method to quantitatively assess the parameters in CWs models, (2) correcting the simulation errors caused by the assumption of Activated Sludge Models (ASMs) and developing a complete biofilm reaction sub-model, (3) simplification of the comprehensive model, and (4) need of emerging pollutants modeling.