The pilot project of CO2 flooding has been implemented in the conglomerate oil reservoir in Xinjiang, China. The target reservoir is characterized by the low-porosity (ranging from 8.53% to 11.02%) and low-permeability (ranging from 3.30 to16.18 mD) parameters. The objective of the pilot CO2 flooding is to overcome the water sensitiveness of the formation. However, the field application of the CO2 injection is complicated by technical difficulties including gas channeling and low sweep efficiency. To control the oil recovery factor, it is critically important to investigate the efficiency of the oil displacement in pores and throats of the formation in the process of water and CO2 flooding. In this paper, in order to quantitatively determine the oil displacement in the pores of different sizes, the authors have studied the process of the water and CO2 flooding in the core samples obtained from the conglomerate oil reservoir in the Xinjiang Oilfield. The displacement efficiency is evaluated by the online nuclear magnetic resonance (NMR) method. Before testing, the samples are saturated with oil till the oil saturation reaches 48.27%. In the process of water flooding, the water content in the large pores reaches 28.70%, the displaced oil is produced mainly from the large pores, and the oil saturation is reduced to 27.78%. In the process of CO2 flooding, the oil displacement efficiency is increased by 17.97%, including an increment of 7.17% for the middle-size pores and 8.61% for the large pores. The large and the middle-size pores contribute to the high oil recovery, while the oil in the small pores is difficult to recover. The ultimate oil saturation is 19.97%, and the overall recovery factor is 60.44%. At the late stage of CO2 flooding, a foam-like structure is developed in the large pores, which is prone to plug the pore throats of the large pores and boost the residual oil displacement in the small pores, thereby further enhancing the oil recovery factor. This study is focused on comprehensive research on the relationship between the pore size and the oil recovery factor in a conglomerate oil reservoir. The results can be useful for guiding the implementation of site-specific EOR technologies.