Insulinoma, owing to the low incidence and small volume of the tumor, is often undiagnosed. The 72-h fast test is centered on diagnosing insulinoma; however, it cannot be performed on outpatients. Our aim was to evaluate the results of a 3-h oral glucose tolerance test (3-h OGTT) for insulinoma diagnosis. Thirty-seven patients with insulinoma were enrolled for comparison with 42 control subjects. All patients underwent 3-h OGTT with measurements of insulin and C-peptide. The secretion ratios of insulin and C-peptide at 1, 2, and 3 h were calculated by comparison with their values at 0 h. We used logistic regression analysis to establish the predictive models and compared the diagnostic efficiency by receiver operating characteristic analysis. The fasting insulin and C-peptide levels of insulinoma patients were both higher; however, the concentrations at 1 h and 2 h were both lower (P < 0.05). The levels at 3 h were not significantly different (P > 0.05). Our final logistic regression model was constructed as follows: logit (P) = 8.305 - 0.441 × insulin 2 h/0 h ratio - 1.679 × C-peptide 1 h/0 h ratio. A cutoff value of > 0.351 showed the highest diagnostic accuracy, with an area under the curve of 0.97, a sensitivity of 86.5%, and a specificity of 95.2%. The 2-h/0-h insulin ratio, as well as the 1-h/0-h C-peptide ratio, has high diagnostic efficiency for insulinoma. The 2-h OGTT can be an alternative test for diagnosing insulinoma in outpatient settings.