In order to operate a CO2 flooding scheme successfully, it is necessary to get accurate information about the reservoir dynamic performance and the fluids injected. Although some numerical simulation studies have been conducted, the complicated drive mechanisms and actual reservoir performance have not been fully understood. Thus, there is a strong industrial need to develop models using different perspectives to provide valuable and complementary insights into the reservoir performance during the CO2 flooding process. The objective of this study is to develop models using material balance equations (MBE) to analyze the field data before and after CO2 injection. After matching the historical field data, the proposed model can be applied to evaluate, monitor and predict the overall reservoir dynamic performance during the CO2 flooding process. To accurately account for the complex displacement process involving compositional effect and multiphase flow, the PVT properties of reservoir fluids and the four-phase fluid relative permeability relationship are integrated into the model. This study has investigated the effects of a number of factors, such as the reservoir pressure, the amount of CO2 injected, the CO2 partition ratios in reservoir fluids, the possibility of the existence of a free CO2 gas cap, the proportion of reservoir fluids contacted by CO2, the oil swelling and the oil relative permeability improvement. The model has been applied to analyze the Weyburn CO2 flooding project. The study has shown that the proposed MBE model is an effective complementary tool to analyze overall reservoir performance in tertiary CO2 recovery processes. The results show that: 1) there exists a free CO2 gas cap under reservoir conditions, even if the reservoir pressure is larger than MMP (minimum miscible pressure) in the Weyburn Field; 2) the CO2 partition ratios in oil, water and gas phases and the proportions of reservoir fluids contacted by CO2 largely affect the drive mechanism and production performance; and 3) the effect of CO2 solubility in water under actual reservoir conditions cannot be neglected. The proposed new model is the first one in developing and applying MBE to evaluate the overall dynamic performance for the CO2 flooding process and a valuable insight into reservoir responses during this process has been achieved.