The rapid development of generative artificial intelligence, such as ChatGPT and Deepseek, has brought new opportunities and challenges to the primary mathematics precision review class, and it is of great significance to explore the roles and modes of teachers, students, and machines in the primary mathematics human-machine collaborative precision review class. This study, guided by the theory of human-machine collaborative education and the integration of ‘teaching-learning-evaluation’, adopting the methods of literature research, theoretical deduction and quasi-experimentation, firstly proposed the general process of “Precise Portrait and Learning Conditions Diagnosis, Deficiencies Check and Methods Refinement, Knowledge Sorting and Network Construction, Precise Practice and Migrating Application, Personalized Improvement and Evaluation and Reflection” in the human-machine collaborative precision review class for primary mathematics. This study then clarified the human-machine synergy mechanism, including precise diagnosis, precise breakthrough, targeted practice, and instant feedback in the primary mathematics precision review classroom, to construct a human-machine synergy precision review classroom mode for primary mathematics. Next, we selected the “Graphics and Geometry Collation and Review” from the second volume of the primary sixth-grade mathematics textbook published by the People’s Education Press as the teaching content and conducted a comparative teaching experiment by selecting A and B classes in the sixth grade of X Primary School in GZ City, with similar test scores and learning abilities. Finally, the effect analysis was conducted in terms of the cultivation of students’ core mathematical literacy, mathematical problem-solving ability, and the effectiveness of mathematical review classes. The results show that the mode can effectively improve students’ core mathematical literacy, mathematical problem-solving ability, and mathematical review class effectiveness.