In the last decades, the Arctic climate has changed dramatically, with the loss of multiyear sea ice one of the clearest consequences. These changes have occurred on relatively rapid timescales, and both accurate short-term Arctic prediction (e.g., 10 days to three months) and climate projection of future Arctic scenarios present ongoing challenges. Here we describe a representation of the Arctic ocean and sea ice in a ultrahigh resolution simulation in which the horizontal grid mesh reduces from 8 km at the equator to 2 km at the poles (UH8to2) for the years 2017-2020. We find the simulation reproduces observed distributions of seasonal sea-ice thickness and concentration realistically, although concentration is biased low in the spring and summer and low biases in thickness are found in the central and eastern basins in the fall. Volume, fresh water, and heat transports through key passages are realistic, lying within observationally determined ranges. Climatological comparisons reveal that the UH8to2 Atlantic Water is shallower, warmer, and saltier than the World Ocean Atlas 2018 climatology for 2005-2017 in the eastern basin. Our analysis suggests that these biases, combined with a lack of stratification in the upper 100 m of the simulated ocean, contribute to the winter biases in modeled sea ice thickness. This relationship between biases in the sea ice and ocean points to a potential positive feedback within the model, illuminating challenges for long term model predictive power in a changing Arctic climate.