Near infrared (NIR) spectroscopy is an instrumental method widely used for rapid and nondestructive detection of internal qualities of agricultural products. Statistical modeling is a very important and difficult process in NIR detection to establish the relationship between spectral information and interested index. Classical multivariate calibration methods such as partial least square regression (PLSR), principle component regression (PCR) and stepwise multi linear regression (SMLR) were often used for modeling. In this study, besides these algorithms, another mixed algorithm was adopted for establishing a nonlinear model of NIR spectra and MT-firmness of pears. The mixed algorithm was combined with SMLR and artificial neural network (ANN). Compared the classical multivariate calibration methods of PLSR, PCR and SMLR, the modeling results using PLSR method of original spectra were much better than the results using derivative spectra and the other two methods: r=0.88, RMSEC=3.79 N of calibration and r=0.83, RMSEP=4.35 N of validation. The mixed algorithm also performed better than SMLR and PCR, but was a bit worse than PLSR: r=0.85, RMSEC=4.15 N of calibration and r=0.82, RMSEP=4.67 N of validation. The results indicated that fruit NIR spectra could be used for MT-firmness prediction when proper algorithm was chosen, however, further study on statistic modeling are still needed to improve the predicting performance.