The aim of this study was to describe and to test the effect of learning model (problem solving learning model oriented toward Balinese local wisdom (PSBLW) Vs Direct Instruction Model (DI)) and type of mathematics problems (open and closed problems) on the ability to solve mathematics problem of the fifth-grade students of elementary in Singaraja Bali. This quasi-experimental research used non-equivalent control group design with pretest and posttest. The data were analyzed with factorial 2 × 2 analysis of covariance (Ancova). The sample consisted of the fifth-grade students of Elementary School with the total of 152 students spread into 4 classes. The sample was selected by cluster random sampling. The data were collected using mathematics problem solving ability test at the 5% significance level (a = 0.05). The statistical analysis was done with the aid of SPSS 16.0 for Windows. The results showed that (1) the ability, may to solve mathematics problems of the students who learned through PSBLW is higher than those who learned through direct instructional model; (2) the students' ability to solve problems facilitated with open mathematics problems was higher than that with closed mathematics problems; and (3) there is effect of interaction between learning model and problem type on problem solving ability. The conclusion is local wisdom-oriented problem-solving learning model effective to improve mathematical problem-solving ability.