Convection‐Permitting Model (CPM) simulations of the Central United States climate for the summer of 2011 are studied to understand the causes of warm biases in 2‐m air temperature (T2m) and related underestimates of precipitation including that from mesoscale convective systems (MCSs). Based on 10 CPM simulations and 9 coarser‐resolution model simulations, we quantify contributions from evaporative fraction (EF) and radiation to the T2m bias with both types of models overestimating T2m largely because they underestimate EF. The performance of CPMs in capturing MCS characteristics (frequency, rainfall, propagation) varies. The pre‐summer precipitation bias has large correlation with mean summertime T2m bias but the relationship between summertime MCS mean rainfall bias and T2m bias is non‐monotonic. Analysis of lifting condensation level deficit and convective available potential energy suggests that models with T2m warm biases and low EF have too dry and stable boundary layers, inhibiting the formation of clouds, precipitation and MCSs. Among the CPMs with differing model formulations (e.g., transpiration, infiltration, cloud macrophysics and microphysics), evidence suggests that altering the land‐surface model is more effective than altering the atmospheric model in reducing T2m biases. These results demonstrate that land‐atmosphere interactions play a very important role in determining the summertime climate of the Central United States.