CO2 flooding is one of the most effective and used methods for enhanced oil recovery (EOR) approaches. The number of CO2 flooding projects has increased rapidly in China and around the world. Compositional simulation is required for evaluating CO2 flooding in EOR operations, especially for miscible or nearly miscible flooding when black-oil simulation is no longer adequate. The simulation method proposed here is a multi-dimensional, three-phase, and compositional modeling approach, which is applicable to both porous and fractured reservoirs. In the model formulation, a generalized multi-continuum approach is adopted to handle flow and transport in naturally fractured reservoirs and the mass flux of each mass component is contributed by advection and diffusion processes. In addition, precipitation of heavy oil components and absorption of CO2 on the solid grains are modeled based on reversible linear or nonlinear isotherms. The governing partial differential equations for conservation of each component are discretized using a finite volume method and the resulting discrete equations are solved fully implicitly by Newton-Raphson iteration. The equation of state (EOS) by Soave-Redlich-Kwong is used to calculate the physical properties of fluids. Research has shown that the flash calculations with EOS in compositional simulation are computationally intensive and may not be reliable at near critical conditions. Therefore, a K-value based approach, improved by Almehaideb et al. (2002), is used for partitioning of oil components and CO2 between oil and CO2 phases. In addition, the laboratory measured oil and CO2 phase compositional data can be used alternatively to account for compositional effect in this model. Two numerical examples are presented to show that the proposed modeling method is efficient for simulation of CO2 flooding processes in EOR operations.