Molecular reconstruction is popular in current oil refinery modeling. It aims to understand the refining process from the molecular level, to predict product properties accurately, to optimize the processes, and to increase the value of each molecule. Molecular reconstruction technique determines the detailed molecular composition of petroleum fractions through obtainable bulk properties and chemical details. In this paper, the existing molecular reconstruction models involving models with a set of predefined deterministic molecules, stochastic reconstruction method, structure-oriented lumping method, molecular type-homologous series matrix method, reconstruction by entropy maximization method, stochastic reconstruction-entropy maximization method and state space representation method are reviewed and compared. The credibility of the simulated composition and the drawbacks of molecular reconstruction technique are also discussed.