The main objective of this study was to test a structural theoretical model of the effects of self-regulated learning on academic stress, subjective well-being, and academic achievement in Secondary Education, considering academic procrastination as a mediator. An additional aim was to explore whether these relationships were moderated by gender and educational level. Participants were 728 students in compulsory and post-compulsory secondary education in a large city in Eastern Spain. Path analysis results indicated that the proposed model showed satisfactory fit, with the three dimensions of self-regulated learning significantly predicting the educational outcomes considered, and that procrastination mediated these relationships. Overall, the model is able to predict 9.8% of the variance of academic stress, 23.1% of students wellbeing, and 14% of academic achievement. Moreover, the multi-group routine revealed no moderation effects due to gender, but educational level moderated two relationships, between self-efficacy and academic achievement and between metacognitive strategies and procrastination. Additionally, supplementary models were tested for three specific subjects (Spanish Language, Foreign Language and Mathematics), which showed an improvement in explained variance, being respectively: 29%, 28% and 27%. Results are discussed in light of previous research and in terms of their impact on educational practice.