This study aimed to compare the mathematical connection ability of students who took mathematics learning with generative learning models and direct learning models. This quasi-experimental study used the posttest-only control group design. The population in the study was 212 people in class X TKJ SMK Informatics Santu Petrus Ruteng spread across six classes, and two classes were chosen as research samples using simple random sampling. The results of the calculation of the normality and homogeneity test as the prerequisite for the t-test showed that the data from the two groups are normally distributed with homogeneous variance. Then the hypothesis test is done using t-test to see the mathematical connection ability. Based on the analysis of the data obtained that the mathematical connection ability of students who use generative learning models is higher than those using direct learning models.