OBJECTIVE： To optimize the extraction technology of total flavonoids from Psidium guajava leaves and investigate the effects of total flavonoids from P. guajava leaves on glucose tolerance in diabetic model mice. METHODS： Based on single factor test， using extraction time， solid-liquid ratio and the amount of ethanol as independent variables， comprehensive scores of the yield of hyperin， quercetin-3-O-β-D-arabinopyranoside， quercetin-3-O-α-L-furanarabino side ethyl acetate and extract as dependent variable， Box-Behnken design-response surface method was used to optimize the reflux extraction process of total flavonoids from P. guajava leaves， and validation test was also conducted. Sixty mice were randomly divided into blank group， model group， metformin group （170 mg/kg）， P. guajava leaves total flavonoids low-dose， medium-dose and high-dose groups （23.41，46.81，93.62 mg/kg， by the content of total flavonoids）， with 10 mice in each group. Except for blank group， diabetic model was induced in those groups， and then model mice were given relevant medicine intragastrically once a day for consecutive 21 d， 6 h after last medication given glucose 2.0 g/kg intragastrically. The levels of blood glucose were determined 0， 0.5， 1.0， 2.0 h after intragastric administration of glucose， and the area under the curve （AUC） of blood glucose was calculated. RESULTS： The optimal technology of P. guajava leaves total flavonoids was as follows as extracting twice， lasting for 1.0 h each time， solid-liquid ratio of 1 ∶ 17， ethanol volume fraction of 56％. In validation test， the contents of hyperin， quercetin-3-O-β-D-arabinopyranoside and quercetin-3-O-α-L-furanarabino side ethyl acetate were 2.57， 3.38， 2.26 mg/g（RSD＜2％，n＝3）； the yield of extract was 25.71％（RSD＝1.19％，n＝3）； average comprehensive score was 96.41 （RSD＝1.34％， n＝3）； relative error was 3.57％ with the predicted value of 99.98. Compared with blank group， AUC of mice were increased significantly in each group； compared with model group， AUC of metformin group， P. guajava leaves total flavonoids medium-dose and high-dose groups were all decreased significantly， with statistical significance （P＜0.05 or P＜0.01）. CONCLUSIONS： The optimal extraction technology is stable， feasible and controllable in quality. Total flavonoids from P. guajava leaves can promote the recovery to normal level of blood glucose.