The traditional Japanese language teaching method is challenging to mobilize students’ enthusiasm, and it is impossible to objectively understand students’ mastery of knowledge points. To address this teaching problem, this paper combines the CRS model with K-Means cluster analysis to construct a precise teaching model for teaching Japanese language in colleges and universities based on the CRS model. Through the CRS model to obtain the students’ mastery of knowledge and differentiated data, the K-Means algorithm is used to analyze the teaching model accurately, so Japanese teachers can make timely adjustments to the classroom teaching situation and improve the quality of teaching. After the model design was completed, a controlled teaching experiment was conducted in two colleges and universities in a province, and an independent square root t-test was performed. The ratio of teachers’ verbal behavior in the experimental class of Japanese language in colleges and universities was 71.58%, which was 18.86% lower than the ratio of teachers’ verbal behavior in the classroom of the control case, and the ratio of students’ verbal behavior was 21.58%, which was 18.36% higher. The average probability of students’ mastery of the Japanese language points studied in the semester was above 0.6. The independent square root t-test for the two classes was less than 0.05, showing a significant difference.