Abstract: Based on the observation and reanalysis data, this study systematically evaluates the global monsoon simulated by the new version of the climate system model FGOALS-g3 by applying moisture budget diagnosis and composite analysis. Moreover, this work analyzes the advantages and disadvantages of the new version when compared with FGOALS-g2. Influences of the air–sea coupling process on the simulated results are discussed by comparing with the corresponding atmospheric component model GAMIL. FGOALS-g3 reasonably reproduces the basic characteristics of the climatology of the global monsoon, including the annual mean precipitation as well as circulation, annual cycle modes, monsoon precipitation intensity, and monsoon region. However, the model underestimates the annual mean precipitation over the land monsoon region, overestimates the annual mean precipitation over the ocean region, and the simulated spring–fall asymmetric mode of the annual cycle is stronger in the tropical monsoon region. The results show that the smaller land monsoon region than the observation in FGOALS-g3 is associated with the weaker vertical moisture advection (especially the thermodynamic term) in summer. For the inter-annual variability, FGOALS-g3 can reproduce the drier pattern of the global monsoon during the El Niño year. However, some biases in precipitation anomalies exist in some monsoon regions. For instance, the precipitation in the west African monsoon region is more than normal, and the precipitation in the southwest Indian Ocean is a dipole anomaly, both of which are inconsistent with the observation. Moreover, the precipitation in the northwest Pacific monsoon region is greater than the observation during the El Niño year. There is no weak convergence center in the upper layer of western Africa in the simulation, and the simulated maritime continent is warmer than observation, resulting in the convective center moving westward during the El Niño year. Compared with FGOALS-g2, FGOALS-g3 improves the simulation of monsoon circulation, inter-annual variability of monsoon precipitation, and monsoon–ENSO relationship. When comparing the coupled and uncoupled simulations, most biases in the coupled model originate from the atmospheric model itself, and the air–sea coupling process partially improves the simulation of precipitation and circulation of the Asian–Australian monsoon region and the tropical Indian Ocean. However, the sea surface temperature bias caused by the coupled process enhances the dry bias of the Indian Peninsula and the wet bias of the tropical Indian Ocean.