This study aims to test the current effectiveness of the application of Problem Based Learning model in Technique Informatics FIKOM Duta Bangsa University of Surakarta on numerical method lesson to improve learning outcomes students. This study using the quasi-experimental technique. Design in this research using posttes control group design with a group of non-equivalent. Population in this study were all students semester VI of engineering study program information FIKOM Duta Bangsa University of Surakarta. In this research data collection was carried out after treatment (posttest). Posttest analyzed data in descriptive with anova to knowing the difference between the experiment and the control group. The method used to normality Wilk Shapiro and the homogeneity variance with Lavene. The prerequisite met then one way Anova with 5% of standard signification. The results of the testing of hypotheses stating that there are differences between students who had learning outcomes by Project Based Learning than students taught by directly learning. As a result, PBL learning, particularly numerical approaches, should be used in lectures. PBL learning is the best, as evidenced by the tests that have been conducted.