Young children are sensitive to both numerical and spatial magnitude cues early in development, but many questions remain about how children's attention to magnitudes relates to their early math achievement. In two studies, we tested three hypotheses related to the flexible attention to magnitudes (FAM) account, which suggests that young children's flexible attention to both numerical and spatial magnitudes is an important predictor of early math success. In Study 1, we recruited 318 preschool-age children (51.5% female; M[subscript age] = 54.7 months, SD[subscript age] = 4.24 months) and assessed them at two time points on a battery of math, executive function, and language measures, including a novel assessment of FAM ability developed for this study. Consistent with our hypotheses, we found that young children had specific difficulty flexibly shifting between numerical and spatial magnitudes, that their FAM ability was related to both their executive function and math skills controlling for covariates, and finally that their FAM ability at the first time point was predictive of their math achievement growth over time controlling for executive function skills and covariates. In Study 2, we recruited 157 preschool-age children (47.4% female; M[subscript age] = 53.36 months, SD[subscript age] = 7.17 months). We replicated the findings of Study 1 and extended them to account for children's non-symbolic numerical magnitude discrimination skills. Implications of the results of these studies for early math activities development are discussed.