The near infrared spectroscopy and chemometrics were employed to discriminate Xihu longjing tea and Zhejiang longjing tea. First of all, the raw spectra were pretreated with standard normal variant SNV function. And then the SNV-pretreated spectra were used as the X variant of PLS-DA; principal component analysis were conducted on the SNV-pretreated spectra, and the first ten scores of principal component were used as the input of least square support vector machine (LSSVM) and radial basis function neural network (RBFNN). The parameters of the relative weight of the regression error (γ) and the kernel parameter of the RBF kernel (δ₂) were optimized with grid research and leave-one-out cross-validation technologies, and the optimized γ and δ₂ were 229.1 and 124.9 respectively. Meanwhile the best numbers of neuron in hidden layer of RBFNN was 27. With comparison the performance of three models, LSSVM showed the best performance. The RMSECV and R₂ were 0 and 1 in calibration set respectively, while RMSEP and R₂ were also 0 and 1 in validation set respectively. As a result, 100% of Zhejiang longjing tea and Xihua longjing tea were classified.