Pulmonary gas exchanges are vital to human health, and disruptions to this process have been associated with many respiratory diseases. Previous gas exchange studies have predominately relied on whole-body testing and theoretical analysis with 1D or static models. However, pulmonary gas exchanges are inherently a dynamic process in 3D spaces with instantaneous interactions between air, blood, and tissue. This study aimed to develop a computational model for oxygen exchange that considered all factors mentioned above. Therefore, an integrated alveolus–membrane–capillary geometry was developed with prescribed rhythmic expansion/contraction. Airflow ventilation, blood perfusion, and oxygen diffusion were simulated using COMSOL. The temporal and spatial distribution of blood flow and oxygen within the capillaries were simulated under varying breathing depths and cardiac outputs. The results showed highly nonuniform blood flow distributions in the capillary network, while the rhythmic oscillation further increased this nonuniformity, leading to stagnant blood flow in the distal vessels. A static alveolus–capillary geometry underestimated perfusion by 11% for normal respirations, and the deviation grew with breathing depth. The rhythmic motion caused a phase lag in the blood flow. The blood PO2 reached equilibrium with the alveolar air after traveling 1/5–1/3 of the capillary network. The time to reach this equilibrium was significantly influenced by the air–blood barrier diffusivity, while it was only slightly affected by the perfusion rate. The computational platform in this study could be instrumental in obtaining refined knowledge of pulmonary O2 exchanges.