Abstract: OBJECTIVE To found a universal near-infrared reflectance (NIR) model, to determine the content of pyrazinamide tablets from different manufacturers. METHODS Using cluster analysis method, diffuse reflectance spectra of 3 samples per concentration of 9 concentrations of simulated samples and 46 samples from 20 manufacturers were divided into calibration and validation groups. Then, the quantitative models were established using partial least squares(PLS) through selection of different wavelength ranges, preprocessing methods and smooth points. RESULTS The root mean square errors of cross validation (RMSECV) was 0.775, and determination coefficients is 99.4%. The root mean square errors of prediction (RMSEP) was 0.962. The correlation value between prediction and true of 27 samples in the validation groups is 99.8%, and the average recovery is 99.9% (RSD=1.26%). All the relative standard deviations of the true value and NIR prediction of 6 true samples from 6 manufacturers were no more than 1.53%. CONCLUSION The established NIR methods can be applied to accurate assay for the pyrazinamide tablets of different manufacturers.