Cultivating high-quality and high-yielding rice varieties is one of the important tasks of current rice breeding, and gel consistency is one of the most important indicators of rice cooking and eating quality. The traditional chemical method for measuring the gel consistency of rice has complicated pretreatment, complicated process, and high reagent consumption. It is difficult to meet the needs of rapid non-destructive testing of the gel consistency for large batches of rice varieties (combinations). The near-infrared spectroscopy analysis technology can quickly, non-destructively, and non-pollutingly analyze. In this study, 137 indica rice varieties (combinations) mainly cultivated or newly bred in southern rice areas were used as the test materials, and their near-infrared spectra were measured by traditional chemical methods to collect their near-infrared spectra to establish near-infrared spectra of southern indica rice. Build an analysis model, and then correct and verify the model. The results showed that the partial least squares method (PLS) was used to establish their respective near infrared analysis models after 20 kinds of mathematical preprocessing and 6 kinds of wavelength bands (or combinations). By comparing the model evaluation indexes, it was determined that smooth pretreatment was the best pretreatment method, and the wavelength band of 1 100 to 1 650 nm was the best modeling wavelength band. Evaluation index of smoothing model: calibration correlation coefficient (R), test correlation coefficient (r), relative analysis error (RPD) were 0.970 0, 0.964 2, and 3.780 5 respectively; wavelength evaluation range: 1 100 to 1 650 nm Model evaluation indicators: R, r, RPD They were 0.969 4, 0.963 8, and 3.758 6 respectively; after smoothing, the best near-infrared analysis model of rice gel consistency was established in the wavelength range of 1 100 to 1 650 nm, and the model evaluation indicators: R, r, and RPD were 0.979 0, 0.974 1, and 4.419 4 respectively; Then used 30 samples to verify the obtained optimal model. It indicated that the absolute error between the near infrared predicted value and the chemical value was 0.198 6~6.502 4 mm, and paired t test showed that p=0.726>0.05, indicating no significant difference between the predicted value and the chemical value. The near-infrared model was feasible for rapid non-destructive testing of rice gel consistency. This study provides technical support for the rapid screening of high-quality rice varieties (combinations) in the early generation of materials and the rapid batch analysis of the gel consistency of rice.