Due to rapidity and accuracy, near-infrared spectroscopy (NIRs) is powerful tool to establish appropriate prediction models with an innovative method for the evaluation of wood properties. In order to reveal mechanical qualities of clonal Chinese fir woods and determine sound prediction models of mechanical properties, four main mechanical properties of six Chinese fir clones (Yang 020, Yang 061, Kaihua 3, Kaihua 13, Daba 8, Kailin 24) were evaluated by NIRs. As a result, Kaihua 13, Kailin 24 and Yang 020 showed good mechanical properties. To estimate mechanical properties with NIRs, different methods should be adopted for different properties. The average spectra of radial section and tangential section combined with multiple scattering correction (MSC) and Savitzky-Golay (S-G) smoothing methods were used to predict the modulus of rupture (MOR) and modulus of elasticity (MOE). By adopting spectra of cross section and taking MSC and S-G smoothing methods for pretreatment, the models of compressive strength parallel to grain could deliver the best results. For wood hardness, the models established with average spectra of three sections and first derivative method were preferred. The correlation coefficients of the prediction models were between 0.84 and 0.90, and those of calibration models were between 0.75 and 0.96.