Simulating the Madden–Julian Oscillation (MJO) and understanding the underlying instabilities driving it remains a significant challenge, despite decades of research. Many of the models suffer from persistent biases, especially in precipitation. To address the gap between current understanding of MJO processes and evaluation and improvement of cumulus parameterizations, scientists from the Department of Energy’s Pacific Northwest National Laboratory simulated two episodes using data collected during the 2011 Atmospheric Radiation Measurement Program MJO Investigation Experiment (AMIE)/DYNAMO field campaign. They used three approaches: a regional model with various cumulus parameterizations (WRF); a regional cloud-permitting model (WRF); and a global variable resolution model (MPAS) centered over the tropical Indian Ocean. They examined the model biases in relationships relevant to existing instability theories of the MJO and quantified their relative contributions to the overall model errors using a linear statistical model. The model simulations captured the observed relationship between moisture saturation fraction and precipitation. However, they found that the precipitation associated with the given saturation fraction was overestimated especially at low saturation fraction values. This bias is found to be a major contributor to the excessive precipitation the models simulate during the suppressed phase of MJO. The spatial and temporal structures of the model-simulated MJO episodes were much improved after accounting for this bias (using a linear statistical model). The remaining biases were strongly correlated with biases in saturation fraction. The excess precipitation bias during the suppressed phase of the MJO episodes was accompanied by excessive column integrated radiative forcing and surface evaporation. They found that the large portion of the bias in evaporation is related to biases in wind speed, which in turn are correlated with those of precipitation. These findings suggest that the precipitation bias sustains itself at least partly by cloud radiative feedbacks and convection-surface wind interactions.