The purpose of this study was to evaluate the performance of atmospheric and coupled simulations from the HadGEM1.2 models family in capturing the interannual variability signal of extreme precipitation events over the Amazon region. The spatial pattern and the thresholds of extreme events for two domains, North Amazon and Amazon (AMN and AMZ) were analyzed comparing the models and CMAP (ClimatePrediction Center - Merged Analysis of Precipitation) data. The composite analysis of precipitation anomalies in interannual scale for austral summer and winter showed, in general, that atmospheric and coupled simulations correctly represented the spatial pattern of such anomalies. Regarding the dry extremes, the spatial pattern was well represented. Overall, the spatial pattern of simulations presented smaller bias in wet extremes. The threshold analysis for wet and dry extremes showed that the simulations represented the interannual scale signal in both domains. The AMZ region presented a smaller bias between simulation and CMAP. Despite the differences, both simulations (atmospheric and coupled) presented similar patterns and therefore represent the interannual scale signal in the two Amazon subdomains.