Several methods exist when seeking to experimentally evaluate the antioxidant properties of a natural bioactive substance. In the case of flavonoids, the methods used are mainly based on the experimental determination of the percentage of inhibition (IC50) or the redox potential (E). In the present work, a prediction study of the redox potential E and the inhibitory concentration LogIC50 was carried out, using the AM1 and HF/6-311G(d,p) method. At the end of this study, three (03) QSPR models were validated and retained, one (01) for the prediction of the redox potential and four (02) for the prediction of the inhibitory concentration : The Redox Prediction Model, developed at the AM1 approximation level, for which 96.43 of the experimental variance is explained by the descriptors : E= -0,29 + 0,22EHomo + 0,11ELumo - 0,05 The Inhibitory Concentration Prediction Models, developed at the AM1 level, for which 96.35⁒ of the experimental variance is explained by the descriptors : LogIC50 = -4,92 + 11,37EHomo + 34,36ELumo + 0,67 The Inhibitory Concentration Prediction Model, developed at the HF/6-311G level (d, p), for which 99.96⁒ of the experimental variance is explained by the descriptors. LogIC50 = 62,40 + 80,25 EHomo - 28,44Elumo + 52,01S - 71,26 η - 6,11μ The development of these QSPR models represents a significant advance in predicting the antioxidant properties of bioactive molecules such as flavonoids based on descriptors calculated by quantum chemical methods.