In this study, a near infrared partial least squares (NIR-PLS) quantitative model, which involved seven adulterants and with methamphetamine purity ranging from 10% to 100%, was established for the first time. Seven adulterants of dimethyl sulfone, isopropyl benzylamine, sucrose, cyclohexylamine, aluminum potassium sulfate, piracetam and ephedrine were most frequently detected in seized methamphetamine samples. High purity methamphetamine and adulterants were mixed to prepare the model samples to make sure the established quantitative model can cover the common adulterant species and purity range of actual seized samples. The characteristic absorption peaks of methamphetamine and adulterants occur in different spectrum range, so the whole spectrum range was used for the PLS modeling. The standard normal variate transformation+first-order derivative (SNV+1D) was proved to be the best spectral pretreatment method. Two separate PLS quantitative models were established to improve the accuracy of the models. The PLS factor, coefficient of determination (R2), root mean square error of cross validation (RMSECV), and root mean square error of prediction (RMSEP) for model 1 was 8, 99.9, 0.8%, and 2.0%, respectively. Model 1 is suitable for high purity methamphetamine samples without adulterant and methamphetamine samples adulterated with dimethyl sulfone, isopropyl benzylamine, sucrose, and cyclohexylamine. The PLS factor, R2, RMSECV, and RMSEP for model 2 was 5, 99.9, 0.8%, and 1.7%, respectively. Model 2 was suitable for methamphetamine samples adulterated with aluminum potassium sulfate, ephedrine, and piracetam. The repeatability and reproducibility for both models were less than 2.1% and 4.0%, respectively. Seventy-two seized methamphetamine samples with purity ranging from 13.9% to 99.4% were used to validate the accuracy of the two models. The average purity determined by liquid chromatography and near infrared spectroscopy was 74.3% and 72.9%, respectively. The t-statistics values were 3.0, which was higher than the significant level of 0.05, so it showed that there was no significant difference between the two methods. Mahalanobis distance and spectral residual were selected as the outlier identification methods. When the Mahalanobis distance value is less than 2, and the spectral residual value is less than 3, the quantitative result is reliable. On the contrary, the quantitative result is unreliable, and the other method is needed for quantitative analysis. The established NIR-PLS method is simple in sample preparation, fast in testing, accurate in quantitative results and high in accuracy. It is suitable for rapid quantitative analysis of methamphetamine in seized samples. The sampling and modeling methods involved in this study are also applicable to other drugs.