Little is known about the distribution and dynamics of macrobenthic communities of the deep Arctic Ocean. The few previous studies report low standing stocks and confirm a gradient with declining biomass from the slopes down to the basins, as commonly reported for deep-sea benthos. In this study, we investigated regional differences of faunal abundance and biomass, and made for the first time ever estimates of deep Arctic community production by using a multi-parameter artificial neural network model. The underlying data set combines data from recent field studies with published and unpublished data from the past 20 years, to analyse the influence of water depth, geographical latitude and sea-ice concentration on Arctic benthic communities.We were able to confirm the previously described negative relationship of macrofauna standing stock with water depth in the Arctic deep sea, while also detecting substantial regional differences. Furthermore, abundance, biomass and production decreased significantly with increasing sea-ice extent (towards higher latitudes) down to values B200 ind m�2,B65 mg C m�2 and B73 mg C m�2 y�1, respectively. In contrast, stations under the seasonal ice zone regime showed much higher standing stock and production (up to 2500 mg C m�2 y�1), even at depths down to 3700 m. We conclude that particle flux is the key factor structuring benthic communities in the deep Arctic Ocean as it explains both the low values in the ice-covered Arctic basins and the higher values in the seasonal ice zone.