Past studies have identified several systematic and persistent biases in the representation of the global water cycle in climate models. These biases reduce our confidence in the model to accurately predict how the water cycle will respond to changes in future forcings and feedbacks. One of the primary goals of the US Department of Energy’s Accelerated Climate Model for Energy (ACME) project is to create an accurate representation of the global water cycle now and in the future. In this study we introduce a set of metrics to diagnose the global water cycle’s mean, spatial distribution, and rain rate distribution characteristics. These metrics are applied to 30-year simulations of an early version of the model run at 1° and 1/4° horizontal resolutions. The metrics identify previously known features, such as an overestimation of the mean precipitation rate, an overestimation of the frequency of drizzle, and an overly rapid cycling of water, which leads to a shortened lifetime of water over the oceans. We analyze cloud and transport features of the model to determine the source of the biases. The impact of observational uncertainty on model evaluation is also explored.