The research is based on the importance of understanding ability in improving the quality of education. Thus, there has to some learning alternatives to support it, that is by using discovery learning models. This model develops an active way of learning by finding the solution for the problem by himself/herself. The Research design used quantitative method with a quasi-experimental design. The population of the research are the fifth grader of elementary school in Rancaekek districk of Bandung Regency with the sample are the fifth grade students of SDN Rancaekek 7 as experiment class and the fifth grade students of SDN Rancaekek 3 as the control class, with the number of 30 students. The purpose of the reseach is to discover the differences of the improvement of mathematical understanding between the students who applied the discovery learning models and those who applied the conventional ones. The analysis technique used are the analysis of quantitative. Based on the results of two average of both class gained a significant of 0,00, the amount is lower than on the significant test, that is 0,05, which means there is a different in ability of mathematical understanding between the students who applied discovery learning model and those who applied the conventional ones. So it can be concluded that the discovery learning model is better than the conventional ones.