Accurate estimation of surface ocean pCO2 is crucial for understanding the ocean’s role in the global carbon cycle and its response to climate change. In this study, we employ a machine learning algorithm to correct the deviations in high-resolution (1/12°) model simulations of surface pCO2 from the INCOIS-BIO-ROMS model (pCO2model) for the period 1980–2019, using available observations (pCO2obs). We train the XGBoost model to generate spatio-temporal deviations (pCO2obs − pCO2model) of pCO2model. The interannually and climatologically varying deviations are then added back to the original model separately, which results in an improved surface pCO2 data product. A comparison of our surface pCO2 data product with moored observations, gridded SOCAT, CMEMS-LSCE-FFNN, and OceanSODA demonstrates an improvement by approximately 40% ± 3.31% in RMSE. Further analysis reveals that adding climatological deviations to pCO2model results in greater improvements than adding interannual deviations. This analysis underscores the ability of machine learning algorithms to enhance the accuracy of model-simulated surface pCO2 outputs.