A coordinated regional climate model (RCM) evaluation and intercomparison project based on observations from a July–October 2014 trans‐Arctic Ocean field experiment (ACSE‐Arctic Clouds during Summer Experiment) is presented. Six state‐of‐the‐art RCMs were constrained with common reanalysis lateral boundary forcing and upper troposphere nudging techniques to explore how the RCMs represented the evolution of the surface energy budget (SEB) components and their relation to cloud properties. We find that the main reasons for the modeled differences in the SEB components are a direct consequence of the RCM treatment of cloud and cloud‐radiative interactions. The RCMs could be separated into groups by their overestimation or underestimation of cloud liquid. While radiative and turbulent heat flux errors were relatively large, they often invoke compensating errors. In addition, having the surface sea‐ice concentrations constrained by the reanalysis or satellite observations limited how errors in the modeled radiative fluxes could affect the SEB and ultimately the surface evolution and its coupling with lower tropospheric mixing and cloud properties. Many of these results are consistent with RCM biases reported in studies over a decade ago. One of the six models was a fully coupled ocean‐ice‐atmosphere model. Despite the biases in overestimating cloud liquid, and associated SEB errors due to too optically thick clouds, its simulations were useful in understanding how the fully coupled system is forced by, and responds to, the SEB evolution. Moving forward, we suggest that development of RCM studies need to consider the fully coupled climate system.