The effects of climate change have been most pronounced in the Arctic. For the future evolution of the Arctic sea-ice cover during the 21st century, climate models show a large scatter between predictions. A new Chinese regional sea-ice–ocean coupled model for the Arctic Ocean was developed. The coupled model was based on the MITgcm ocean circulation model and classical Hibler79 type two category thermodynamics–dynamics sea-ice model. The sea-ice dynamics were considered based on viscous–plastic (VP) and elastic–viscous–plastic (EVP) rheology. The sea-ice thermodynamics were considered based on Winton three-layer models. A high-resolution thermodynamic snow-ice model (HIGHTSI) was introduced as an independent module to improve the sea-ice surface heat balance. A detailed configuration of coupled model has been introduced. Special attention has been paid to the model grid setup, subgrid parameterization, ice–ocean coupling and open boundary treatment. The coupled model was then applied for two climatology simulations. Firstly the model was integrated up to 50 years driven by 10 year (1992–2002) average NCAR/NCEP reanalysis data. The second model run was carried out for the period 1992–2007. The atmospheric forcing was daily NCAR/NCEP reanalysis and the initial condition was the result of first model run for the 50th year. The climatology simulation captured the general pattern of the sea-ice thickness distribution of the Arctic, i.e. the thickest sea ice is situated around the Canadian Archipelago and the north coast of Greenland. For the second model run, the modeled September sea-ice extent from 1992 to 2007 was highly correlated with the observations, with a linear correlation coefficient of 0.86. The minimum of the Arctic sea-ice extent in the September of 2007 was unprecedented. The modeled sea-ice area and extent for this minimum was overestimated relative to the observations. However, model run captured the general pattern of the sea-ice retreat.