There have been many models of decline curve analysis in estimating reserves and production in petroleum reservoirs. It is difficult to evaluate the models because the true value of reserve in oil reservoirs is usually unknown. Previously, comparison and analysis among limited models has been made using spontaneous imbibition data. In this study, an approach to using core flooding experimental data was proposed to verify and evaluate the decline curve analysis models. Four representative models (exponential, hyperbolic, harmonic, and the mechanistic Li-Horne models) were chosen to predict and match the oil production in low permeability core samples and oil reservoirs. A systematic comparison among the four models was conducted by using the sum of squared residuals, regression coefficient, and the true values of reserve (for core samples) as the evaluation parameters. Both experimental and reservoir data from a low permeability oil reservoir with great heterogeneity were used. It was found that it was difficult to appraise the goodness of the decline curve analysis models by using the sum of squared residuals and regression coefficient as judging parameters in some cases. The small values of the sum of squared residuals or the great values of the regression coefficient could not guarantee small error in the predicted reserves. It may be a suitable approach to using experimental data of oil production in core samples with different permeability to evaluate decline curve analysis models. The results demonstrated that the Li-Horne model and the exponential model predicted the reserves satisfactorily for experimental data of water flooding in the core samples with low permeabilities.