Arctic climate is difficult to model accurately due in part to challenges with modeling clouds. In particular, while Arctic clouds are often in a mixed-phase state despite persistent subfreezing temperatures, many models including CAM5 predict insufficient liquid in Arctic clouds. These errors propagate into errors with surface and top-of-atmosphere radiative fluxes, adversely affecting confidence in model projections of Arctic climate change. Here we conduct a thorough comparison of Arctic cloud properties and surface and top-of-atmosphere radiative fluxes between CAM5 and observations, and explore causes of model biases. Two modified mixed-phase ice nucleation schemes are explored in CAM5 and their impacts on Arctic clouds and radiative fluxes are evaluated.