Abstract In order to explore the origin and significance of music education reform and talent cultivation in the new era, this paper constructs a new music curriculum evaluation model based on SVM. Firstly, we explore the path of music education reform and the goal of music talent cultivation by taking the trend of Internet of everything as the origin of the proposed “high-quality courses” for music in universities. Then, we analyzed that the recognition accuracy of SVM is significantly better than most machine learning methods in solving high-dimensional small sample data, so we established a new music curriculum teaching evaluation model on this basis. The average score of each index of the new course teaching effect is 74, and the evaluation grade is better, obtained by establishing the evaluation system of the new music course of SVM. In the evaluation index, the CVI indexes were all kept in the range of 0.82 to 0.98, which was higher than the standard 0.78 of the CVI index, and the Kappa value of each evaluation index was higher than 0.86. The attendance rate in college teaching rose from 78% to 100%, the ability to master music knowledge rose from 67% to 93%, the ability to apply knowledge rose from 76% to 92%, and the average score on weekly tests rose from 86 to 96. Finally, it is concluded that the Internet-based music education reform approach can promote the development of music education careers and cultivate the important means of all-around talents.