The determination of mathematical models for predicting the mass of kiwi fruits from their dimensions and projected areas was done. The research was conducted on 200 observations of two Iranian varieties of kiwi fruit (Abbott and Hayward). The physical characteristics measured included mass, volume, dimensions and projected areas perpendicular to major diameters. Maximum, mean, minimum values, standard deviation, coefficient of variation, bulk density (BD), geometric mean diameter (GM), and percent sphericity (SPH) of all of the parameters were determined. Models were divided into three classifications: 1 – single and multiple variable regression models of kiwi fruit dimensions (dimension models), 2 – single and multiple variable regression models of projected areas (projected area models), 3 – estimation of kiwi fruit shape, ellipsoid or spheroid based on volume (models based on volume). Among single variable models, 1st classification, second diameter model of kiwi fruit had maximum coefficient of determination, R2 = 0.78 for all observations. Among the second classification, mass models of projected areas had a nonlinear relationship with R2 = 0.97. Third classification models had the highest performance followed by the second classification and, respectively, the first, with R2 approaching unity.