Based on the comparison with observation and reanalysis data, the study evaluated the performance of the Atmospheric Model Intercomparison Project (AMIP) and the Coupled Model Intercomparison Project (CMIP) historical experiments of IAP/LASG climate system model FGOALS-g3 in simulating climatology and interannual variability of JAS (July–August–September) seasonal-mean North Africa Summer Monsoon (NASM) and Sahel precipitation, and explained the bias by moisture budget and regression analysis, investigating the influence of ocean-atmosphere coupling by comparing AMIP and Historical. The results showed that both Historical and AMIP experiments underestimated precipitation, simulating weaker south west monsoon winds and a further south rainfall position. The pattern correlation coefficients of precipitation in Sahel and North Africa monsoon region simulated by AMIP are 0.80 and 0.62, respectively, which are larger than those of Historical, and the corresponding root mean square errors are 2.58 and 3.23, which are smaller than those of historical experiment, indicating that the deviation of AMIP is smaller than that of Historical. Considering the moisture budget diagnosis, Historical and AMIP both underestimated the water vapor convergence over NASM region, estimating less vertical moisture advection and evaporation and more horizontal moisture advection than observation, which led to dry biases. In terms of interannual variability, the observation shows that North Africa summer monsoon rainfall is negatively correlated with ENSO. AMIP can reproduce the ENSO-NASM negative relationship, which is stronger than observation. However, Historical cannot reasonably simulate the relationship at interannual time scale. AMIP overestimates the circulation response of ENSO, including descending anomalies, weakened tropical easterly jet and decreased low-level monsoon over North Africa, which contributes to the stronger precipitation negative anomaly. In contrast, Historical underestimates the above ENSO-related response, resulting in feeble precipitation negative anomaly. Moisture budget analysis indicated that vertical moisture advection anomalies, especially the dynamic term of vertical moisture advection anomalies, dominated the ENSO-NASM negative relationship. AMIP is coincided with observation, but it overestimates the above term, which leads to stronger negative rainfall anomalies. While Historical overestimates horizontal advection and vertical thermodynamic anomalies, which indicates that horizontal advection anomalies cause the inhibited simulation of ENSO-NASM negative relationship.