Abstract: The internal quality indicators such as soluble solids content (SSC), total acid (TA), pH, firmness index (FI) and moisture content (MC) of red globe grapes affect the taste and quality of the fruit directly. And they are also criterion for maturity. In order to obtain the internal quality indicators showed above quickly and avoid unnecessary inspection losses, a new non-destructive detection model for red globe grapes was proposed in this paper. The near-infrared spectral information of 360 samples was collected using the Antaris II near-infrared spectrometer for red globe grapes in the growing period. The collected spectral information was pre-processed by algorithms such as SNV and then modelled to determine the optimal spectral pre-processing method. The characteristic wavelengths of the spectral information were extracted by dimensionality reduction algorithms. Finally, the detection models for SSC, TA, pH, FI and MC of red grapes were established respectively based on Partial Least Squares Regression (PLSR) algorithm. For SSC and TA the optimal detection model was SG-CARS-SPA-PLSR, for pH the optimal detection model was MA-CARS-SPA-PLSR, and for FI and MC the optimal detection model was SG-CARS-PLSR. The correlation coefficients (R_p) of the optimal PLSR models established of the prediction sets for red globe grape SSC, TA, pH, FI and MC were 0.9787, 0.9811, 0.9870, 0.9568 and 0.9329 respectively, and the residual prediction deviations(RPD) were 4.8637, 4.9006, 6.0939, 3.4453 and 2.5825 respectively, indicating that the above models had high detection accuracy. The models established in this paper would provide a reliable method for the detection of the internal quality of red globe grapes.ity of red globe grapes.