Two feeding trials using different commercial feeds and different fish sizes were conducted to compare model performance of various regression models on the estimation of optimum feeding rates (OFR; % BW/day) for juvenile olive flounder. Six feeding rates ranging from 1% to 11% in trial I and another six feeding rates ranging from 1% to 6% in trial II were tested. Final body weight, weight gain, feed conversion ratio, condition factor, nutrient gains, and whole‐body lipid and energy contents were significantly influenced by the feeding rates in both trials. The overall responses shown in trial I were generally comparable to those observed in trial II. In trial I, the OFR was estimated to be 7.3%, 5.7%, 9.8% and 10.2% by the one‐slope straight broken‐line, two‐slope straight broken‐line, quadratic broken‐line and quadratic models, respectively. In trial II, the OFR was estimated to be 4.8%, 4.6%, 8% and 8% by the one‐slope broken‐line, two‐slope broken‐line, quadratic broken‐line and quadratic models, respectively. Based on the model selection criteria (R2adj and AICc), the quadratic broken‐line model was selected as the best‐fit model in both trials. This model that describes the dose–response relationship best among the tested models can be recommended when determining an OFR.