Grounded in expectancy-value theory, this study examined whether specific academic emotions mediate and moderate the relationship between motivational beliefs (X) and academic performance (Y) in science during adolescence. The sample comprised 340 Uruguayan secondary school students (M = 16.5 years, SD = 0.96; 42.6% female) from five public and private institutions. After removing seven outliers and twelve cases with missing data on the outcome variable, analyses were conducted with 321 cases. A cross-sectional correlational design was used. Analyses were conducted using the PROCESS macro for SPSS (Model 5, version 4.2), which allows estimation of an indirect effect (via a mediator, M) and a moderation of the direct effect by a moderator (W). Motivational beliefs (X) were assessed through the interaction of adapted items assessing expectancy of success and task value. Achievement emotions were measured using a shortened version of the Achievement Emotions Questionnaire, grouped into two dimensions: academic disengagement emotions (M, e.g., hopelessness, boredom) and learning-facilitating emotions (W, e.g., enjoyment, pride). Academic performance (Y) was indexed by one of the final evaluation grades in a Physics course. Motivational beliefs and learning-facilitating emotions were mean-centered prior to analysis. Heteroscedasticity-consistent standard errors were used (HC3 estimator; Davidson & MacKinnon, 1993) to account for observed violations of normality and heteroscedasticity. Results showed a significant indirect effect of motivational beliefs on academic performance through disengagement emotions (ab = 0.015, 95% CI [0.002, 0.029]), suggesting that lower academic disengagement partially explains the positive association between motivational beliefs and performance. The direct effect of motivational beliefs on performance remained significant (B = 0.065, p = .013) and was moderated by learning-facilitating emotions (interaction term: B = -0.038, p = .047). Conditional effects analysis showed that motivational beliefs significantly predicted performance at low and medium levels of learning-facilitating emotions, but not at high levels.