Abstract: In this study, according to observation and reanalysis data, we evaluated the performance of the South Asian summer monsoon (SASM) using the FGOALS-g3 climate system model. We focused on the differences between FGOALS-g3 and FGOALS-g2, and the coupled model and atmospheric model. The results showed that compared with FGOALS-g2, FGOALS-g3 better simulated climatological Indian Ocean trade winds and the sinking branch of the Walker circulation during El Niño, owing to the change in local sea surface temperature (SST). The FGOALS-g3 model showed that systematic cold biases persisted in the middle and upper troposphere, which reduced the meridional temperature gradient and weakened SASM, leading to biases in descending motion and moisture divergence, and dry biases occurred over the terrestrial monsoon trough. Meanwhile, the negative correlation between El Niño—Southern Oscillation and Indian summer rainfall captured by FGOALS-g3 was weaker than the observation, owing to the weaker descending motion caused by SST biases. The results also showed that climatological SST biases induced by air–sea interactions compensated the wet biases in the SASM region through the change in atmospheric circulation and water vapor transportation. At an interannual timescale, the inclusion of the negative feedback process of SST–precipitation–cloud shortwave radiation in the coupled model effectively improved the bias intensity of rainfall and atmospheric circulation simulated by the atmospheric model; however, the westward biases of the sinking branch of the Walker Circulation caused by SST biases in the coupled model led to greater wet biases in the Indian Peninsula.