<p>Artificial intelligence (AI) education empowers teachers to enhance the educational process. Although conventional face-to-face or fully online training methods each have their strengths, they do not fully address challenges such as the rapid pace of AI advancements, differences in teachers’ ability to grasp AI knowledge, and the need for flexible scheduling. The hybrid-flexible (HyFlex) teaching approach integrates the benefits of both methods and offers new possibilities for AI training for teachers. However, in-depth research on the practical impacts and implementation pathways for this model in AI teacher training is limited. To address this gap, we explored three key questions: (1) how to implement HyFlex in AI teacher training; (2) teachers’ engagement and satisfaction in AI training and the corresponding learning outcomes; and (3) how engagement influences teachers’ satisfaction and AI competence. Specifically, we developed a HyFlex teacher training model aimed at enhancing teachers’ AI competence and evaluated its effectiveness in a practical application. The results indicate that (1) The flexibility of HyFlex resulted in higher teacher engagement in the AI course training process, contributing to higher learning outcomes and higher teacher attitudes towards the HyFlex model. (2) The flexibility of learning styles and times in HyFlex implementation and the reusability of AI learning materials help to accommodate teachers from different educational backgrounds. (3) Emotional engagement significantly influences training satisfaction and AI competence in AI teacher training. This study provides valuable insights for future practical applications in AI teacher training and the dissemination of the HyFlex model.</p>