Our study is among the first to provide a comprehensive review of cross-national patterns of gender differences in various STEM-related constructs--achievement, beliefs, attitudes, aspirations, and participation, concerning country-level gender equality. We complement our review with empirical analyses utilizing rigorous methodologies and richer datasets from individual and country levels. Specifically, we examine gender differences in relative strength measures (e.g., strength in science relative to math and reading) and STEM aspirations and graduation, using PISA 2015 and PISA 2018 data from 78 countries/regions (N = 941,475). Our analysis corroborates our literature review, indicating that support for both the gender stratification hypothesis and the gender equality paradox (i.e., whether gender gaps favoring male students are smaller or larger in more gender-equal countries) is generally inconsistent and weak. Various factors contribute to this inconsistency, including specific outlier countries, different years of data collection, diverse data sources, a range of composite and domain-specific measures of gender equality, and statistical models. Our study also introduces a robust statistical model to compare performances in three subjects and evaluate the predictive power of relative strength measures for STEM aspirations at the student level. Our analyses reveal that general academic achievement and math achievement relative to reading are key predictors of STEM aspirations, compared with science achievement relative to math and reading. By juxtaposing both levels of analysis, our findings offer a more nuanced understanding of gender differences in decision-making processes that lead to careers in STEM-related fields.