To explore the feasibility of using near-infrared reflectance spectroscopy(NIRS) to evaluate alfalfa quality rapidly in the field and try to find the appropriate machine and sample preparation method, the representative population of 170 fresh alfalfa samples collected from different regions with different stages and different cuts were scanned by a portable NIRS spectrometer （1 100～1 800 nm）. This is the first time to build models of fresh alfalfa to rapidly estimate quality in the field for harvesting in time. The calibrations of dry matter(DM), crude protein(CP), neutral detergent fiber(NDF) and acid detergent fiber (ADF)were developed through the partial least squares regression(PLS). The determination coefficients of cross-validation (R2CV) were 0.831 4, 0.597 9, 0.803 6, 0.786 1 for DM, CP, NDF, ADF, respectively; the root mean standard error of cross-validation(RMSECV) were 1.241 1, 0.261 4, 0.990 3, 0.830 6; The determination coefficients of validation (R2V) were 0.815 0, 0.401 1, 0.784 9, 0.752 1 and the root mean standard errors of validation(RMSEP）were 1.06, 0.31, 0.95, 0.80 for DM, CP, NDF, ADF, respectively. For fresh alfalfa ,the calibration of DM, NDF, ADF can do rough quantitative analysis but the CP’s calibration is failed. however, as CP in alfalfa hay is enough for animal and the DM, NDF and ADF is the crucial indicator for evaluating havest time, the model of DM, NDF and ADF can be used for evaluating the alfalfa quality rapidly in the field.