The ORCHIDEE land surface model has recently been updated to improve the representation of high‐latitude environments. The model now includes improved soil thermodynamics and the representation of permafrost physical processes (soil thawing and freezing), as well as a new snow model to improve the representation of the seasonal evolution of the snow pack and the resulting insulation effects. The model was evaluated against data from the experimental sites of the WSibIso‐Megagrant project (www.wsibiso.ru). ORCHIDEE was applied in stand‐alone mode, on two experimental sites located in the Yamal Peninsula in the northwestern part of Siberia. These sites are representative of circumpolar‐Arctic tundra environments and differ by their respective fractions of shrub/tree cover and soil type. After performing a global sensitivity analysis to identify those parameters that have most influence on the simulation of energy and water transfers, the model was calibrated at local scale and evaluated against in situ measurements (vertical profiles of soil temperature and moisture, as well as active layer thickness) acquired during summer 2012. The results show how sensitivity analysis can identify the dominant processes and thereby reduce the parameter space for the calibration process. We also discuss the model performance at simulating the soil temperature and water content (i.e., energy and water transfers in the soil‐vegetation‐atmosphere continuum) and the contribution of the vertical discretization of the hydrothermal properties. This work clearly shows, at least at the two sites used for validation, that the new ORCHIDEE vertical discretization can represent the water and heat transfers through complex cryogenic Arctic soils—soils which present multiple horizons sometimes with peat inclusions. The improved model allows us to prescribe the vertical heterogeneity of the soil hydrothermal properties. Key Points The ORCHIDEE land surface model is calibrated and evaluated at local scale for two sites typical of Arctic environments (forest and shrub‐tundra) A global sensitivity analysis identified the parameters having most influence on water and energy flux simulation Model performance and improvements provided by the vertical discretization of the soil hydrothermal properties are highlighted