This study performed latent profile analysis from more than 4000 saliva cortisol samples collected from children at the ages of 2 (T1), 3.5 (T2), and 5 years (T3). Three clearly different cortisol profiles were identified. The largest group at every age point was the Low/Regular latent profile, in which the cortisol slopes followed typical diurnal variation. A smaller proportion of the children belonged to the latent profile with relatively Low/Flat slope, and a minority belonged to the High/Fluctuating latent group, where the overall cortisol values and variations between the slopes were clearly higher than in the other groups. Most of the children who belonged to the High/Fluctuating group were cared for at home, they had higher temperamental surgency and their mothers had more depressive symptoms than in the other latent profile groups. However, only moderate intraindividual stability in diurnal cortisol profiles was observed across the follow-up period. On average, half of the children moved between the groups from T1 to T3. Neither child temperament, social competence, nor sex explained the stability or movement between the groups across age. Variations in cortisol profiles may be caused by the child’s age, and diurnal cortisol rhythm becomes more regular along with development. Methodological issues regarding saliva cortisol research in young children are discussed. Also, more longitudinal research is needed to clarify mechanisms between environmental as well as individual factors and possible dysregulation in a child’s HPA axis functioning.