Identifying the underpinnings of mathematics proficiency is relevant for all societies. A growing literature supports a relation between executive function (EF) and mathematics across a wide age range, but causal links are not well understood. In the current study, typically developing preschool children (N = 104) were randomly assigned to one of four training conditions: EF, Number, EF + Number, or an active Control. They participated in three brief training sessions and pretest and posttest sessions measuring EF and mathematics skills. EF training improved EF skills on a task similar to the training but did not extend to an untrained EF task. In addition, the EF training improved number skills but not general mathematics skills. The EF + Number training improved number and general mathematics skills but not EF skills. The EF + Number training did not yield significantly greater benefits for EF and mathematics beyond other training conditions. Finally, differential training effects emerged, such that children with lower pretest EF skills had greater EF benefits on only the trained EF skill. In addition, children from lower versus higher socioeconomic households had greater gains in numerical skills following EF training. No training condition improved verbal knowledge, suggesting that results were specific to the targeted skills. These results extend prior findings on the effectiveness of improving EF and mathematical skills through short-term trainings during early childhood.