This study seeks to identify learning models that affect students' mathematics learning outcomes. Students at SMPN 3 Paron have poor or low learning outcomes, especially in the field of Harmony in the Triangle, because the indicators of the learning process in the school have not been optimal. One of the learning models that can be used is the Teams Games Tournament (TGT) model with a contextual approach. This quantitative study used Postest only Control Design, with three samples selected using saturated sampling, one class as the experimental class and one class as the control class, and using one class, namely class VIII B as the test class. The samples have been distributed normally and homogeneousl. Data analysis using the T. Teams Games Tournament Test produced an average of 42.44, the conventional learning model produced an average of 62.39. Students who were taught using the Teams Games Tournament (TGT) model with a contextual approach obtained a better average and outperformed students who were taught using the convenive model or only the TGT model, as the calculation of this study obtained (Thitung = 3.024 is greater than Ttable = 2.120).