<p>This study investigated latent profiles of mathematical literacy among Turkish students in the Programme for International Student Assessment (PISA) 2022 and identified key predictors of profile membership across ten plausible values (PVs). Latent profile analysis, conducted on four subdimensions, revealed two distinct profiles: high and low mathematical literacy. To determine the most influential predictors, Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression was applied to 70% of the data (training sets), addressing multicollinearity and selecting variables among 28 theoretically grounded predictors using the lambda.1se criterion. Analyses repeated across the 10 PV-based training sets showed that 14 variables were consistently selected by LASSO, all of which were significant in standard logistic regression models. The results showed that mathematics preference, mathematics self-efficacy, gender (male), socioeconomic status, information and communication technology (ICT) use in mathematics lessons, familiarity with mathematical concepts, online information practices, growth mindset, and expected occupational status positively predicted membership in the high mathematical literacy profile. Conversely, long-term absenteeism, weekday leisure-time ICT use, school-related ICT use, and ICT use for feedback and support negatively predicted high-profile membership. Collectively, these variables explained 37% of the variance in profile membership. Receiver operating characteristic (ROC) analyses based on the remaining 30% of the data (test sets) demonstrated strong predictive performance, with area under the curve (AUC) values ranging from 0.80 to 0.83 across PVs. Cross-validation confirmed the model’s robustness and produced a concise set of predictors relevant to mathematical literacy profiles. The study discussed implications, limitations, and recommendations for research, practice, and policy.</p>