The combination of spatial smoothing and asymptotic analysis allows reduction of computationally expensive 3D fuel cell models to 2D without sacrificing leading-order physics. This paper investigates, demonstrates, and verifies the spatial smoothing and asymptotic reduction of a 3D non-isothermal model for the planar solid oxide fuel cell and stack. Particularly, spatially-smoothed energy equations are developed subject to LTE (local thermal equilibrium) and LTNE (local thermal non-equilibrium) conditions in the flow field consisting of parallel plain channels and solid ribs. The selection of either the LTE or LTNE set for use depends on the temperature difference between the gas flow in channels and the ribs. The reduced models agree well with the 3D counterpart in view of the quantified loss of information due to reduction, while the computational cost is reduced by more than three orders of magnitude. The present methodology is generic and can be applied to other types of fuel cells which are slender in shape and equipped with parallel channels. The reduced model allows statistical sensitivity analysis of cell/stack performance with respect to modeling parameters in a large sample size at computational cost that is not prohibitive. (C) 2014 Elsevier Ltd. All rights reserved.