This paper aims to propose a stress–strain model for the stub column behavior of cold-formed steel non-slender elliptical hollow sections (EHS) under axial compression by considering the post-local buckling behavior and taking into account the effects of local imperfection, material nonlinearity and residual stresses. The proposed model is applicable to the sections of both high strength and normal grade steels; and is mainly defined by four key parameters, including the 0.2% proof stress, the local buckling stress and buckling strain, as well as the post-peak residual strength of a stub column. Equations are also proposed to express these key parameters in terms of the basic input cross-section parameters and virgin material properties. In order to determine these key parameters, existing test and numerical results on the structural behavior of cold-formed steel EHS stub columns were collected and analyzed, which covered a wide range of steel grades (up to 960 MPa) and section sizes. Existing design methods, which have not yet been calibrated for cold-formed high strength steel EHS, were first evaluated by comparing their predicted stub column strengths (i.e., local buckling stresses) with the collected test and numerical data. The Continuous Strength Method(CSM) was found to provide better predictions than other examined design methods. Modified CSM base curves were then developed to further improve the accuracy of the existing CSM for both the local buckling strains and design strengths of cold-formed EHS with steel grades up to 960 MPa. In addition, a theoretical framework is proposed to develop the semi-empirical equations for both the 0.2% proof stresses and local buckling stresses of stub columns by incorporating the modified CSM base curves, whereas one curve based on the conventional equivalent diameter and one curve based on equivalent diameter of the four-arc approximation. A predictive expression of the post-peak residual strengths is also presented. Predictions from the proposed stress–strain model were in a good agreement with the collected test and numerical results, which can demonstrate its accuracy and validit.