Youths' subjective well-being (SWB) is theorized to be an integral element of school success. However, little is known about the relative predictive power of different SWB indicators on educational outcomes. Thus, we investigated youths' global and school-specific SWB as predictors of multidimensional educational performance. Participants were adolescent students (N = 893) in Grades 9-12 within one district and one school in the United States. Regression models were run using adolescent self-reported survey data and school-reported student sociodemographics and outcomes. Global SWB predictors were self-reports of life satisfaction and positive affect; school-specific SWB predictors were self-reports of joy of learning, school connectedness, educational purpose, and academic efficacy. The educational outcomes of interests were end-of-term grades in English language arts, math, and overall coursework, as well absences and suspension. Our models accounted for several student sociodemographic covariates that might affect academic performance. Results showed that, when general SWB predictors were modeled alone, life satisfaction was a consistently meaningful predictor of all educational outcomes. However, when school-specific SWB predictors were added to the models, the predictive power of life satisfaction attenuated, and academic efficacy emerged as the strongest and most consistent predictor of all educational outcomes. Interestingly, joy of learning, school connectedness, and educational purpose did not show meaningful predictive power across educational outcomes. Implications for research and practice are discussed.