This study has a formulation of the problem: (1) whether there is a difference in the initial ability of mathematical connections of control class students with experimental classes; (2) whether the mathematical connection ability of students whose learning uses the Auditory Intellectually Repetition (AIR) model is better than that of students whose learning uses conventional models; (3) whether there is an increase in the mathematical connection ability of students given the Auditory Intellectually Repetition (AIR) learning model with the conventional model. This study is an experimental study with saturated sampling techniques. The results in this study: (1) hypothesis tests on the pretest answer results showed that H0 was accepted and H1 was rejected with an average score of the experimental class pretest score of 33.73 and the control class of 33.27; (2) the results of the hypothesis test on the posttest yang answer showed that H0 was rejected and H1 was received with an experimental class posttest average score of 86.91 and an average control class score of 81.27; (3) the hypothesis test results on the gain value showed that H0 was rejected and H1 was received with an average score of the experimental class gain value of 0.81 and an average score of the control class of 0.73. Researchers can conclude that (1) there is no difference in the initial ability of mathematical connections of control class students with experimental classes; (2) the mathematical connection ability of students whose learning using the Auditory Intellectually Repetition (AIR) model is better than students whose learning uses conventional models; (3) improving the mathematical connection ability of students given the Auditory Intellectually Repetition (AIR) learning model better than students given the conventional model.