In order to improve the detecting accuracy rating and stability of total flavonoids content in ginkgo leaves by near infrared spectroscopy technique, a precision model was established by selecting efficient spectral regions combined with different partial least squares (PLS) selecting wavelength regions methods. Three improved partial least squares (PLS) methods, including interval partial least squares (iPLS) selecting wavelength regions method, backward interval partial least squares (biPLS) selecting wavelength regions method and synergy interval partial least squares (siPLS) selecting wavelength regions method were used to find the most informative ranges and build models with better predictive flavonoids content in ginkgo leaves at first. And then the models were compared with PLS model which was developed on the whole wavelength range 4000~8000cm-1. Results showed that the models built by the three improved PLS methods had higher predictive ability than that of PLS method. The optimal model was the one that obtained by siPLS selecting wavelength regions method and it separated the whole spectra into 21 intervals and combined two intervals including interval 7 and interval 12, the RMSECV and RMSEP were 2.9500 and 3.000, calibration and the prediction correlation coefficient were 0.9384 and 0.9437. The conclusion is siPLS method can accurately and rapidly predict flavonoids content in ginkgo leaves.