This study aims to systematically explore the empirical impact of a physical education teaching model based on the deep integration of DeepSeek big language model and multi-dimensional generative intelligent tools (such as image and video generation tools) on the teaching effectiveness of middle school students. The study adopted the "non equivalent control group pre-test design" in quasi experimental research, selecting two parallel classes in the second year of junior high school at Jinben Middle School, divided into an experimental class (n=45) and a control class (n=45), for a one semester (16 weeks) teaching intervention. The experimental class implements the "DeepSeek+" teaching mode with "personalized pre class preparation, human-machine collaboration during class, and precise extension after class" as the core, while the control class maintains the traditional teaching mode. This paper collect data through various tools such as standardized sports skill tests, sports theory knowledge papers, sports learning interest scales, deep learning process questionnaires, and semi-structured interviews. The research results showed that after controlling for pre-test differences, the experimental class students had significantly higher post test scores in multiple dimensions, including motor skill mastery level (p<0.01), sports theory score (p<0.001), sports learning interest (p<0.01), and deep learning ability (such as critical thinking, communication and collaboration, p<0.05), compared to the control class. Qualitative interview data further revealed that this model effectively enhances students' self-efficacy and intrinsic motivation by providing immediate and accurate feedback, implementing personalized learning paths, and creating high immersion learning contexts. The "DeepSeek+" physical education teaching model can comprehensively and multi-dimensional improve teaching effectiveness, not only in terms of knowledge and skills, but also in terms of stimulating interest and promoting the development of higher-order thinking abilities, showing great potential. This study provides solid empirical evidence for the deep integration of generative artificial intelligence and physical education, and puts forward practical suggestions for the construction of future smart physical education classrooms.

Luojun Li. Empirical Study on the Teaching Effectiveness of Physical Education Classroom Driven by DeepSeek and Multi Generative Intelligent Tools. Frontiers in Sport Research (2025), Vol. 7, Issue 5: 95-101. https://doi.org/10.25236/FSR.2025.070515.

[1] Zhu Z T, Hu J. The Practical Logic and Promotion Strategy of Digital Transformation in Education [J]. China Electronic Education, 2022, (5): 1-10.

[2] Li Y. Research on Artificial Intelligence based Physical Education Action Recognition and Evaluation System [J]. Sports Science, 2021, 41 (8): 75-82.

[3] Wang Z J. How Generative Artificial Intelligence Reshaps Education: Current Status, Path, and Challenges [J]. China Electronic Education, 2023, (10): 1-10.

[4] Kirk D. Educational Value and Models-Based Practice in Physical Education[J]. Educational Philosophy and Theory, 2013,45(9), 973-986. 

[5] Ryan,R.M.,&Deci, E.L. Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being[J]. American Psychologist,2000, 55(1),68-78. 

[6] Nelson Laird, T. F., et al. Deeply Affecting First-Year Students' Thinking: Deep Approaches to Learning and Three Dimensions of Cognitive Development[J]. The Journal of Higher Education, 2014, 85(3), 402-432. 

[7] Yin Z H. The dilemma and solution of sports learning evaluation under the background of core literacy [J]. Sports Journal, 2020, 27 (1): 121-127.