This study tested which of two crucial mechanisms of working memory (WM): attention control, consisting of focusing attention on the proper task-set as well as blocking distraction, and the active buffer capacity, related to the number of chunks that can be actively maintained, plays a more important role in WM’s contribution to fluid intelligence. In the first study, the antisaccade task was used, the standard measure of attention control, in a modified variant which resulted in scores less sensitive to individual differences in the active buffer capacity, in comparison to the standard variant. In effect, attention control became a weak predictor of Gf, explaining less than one third of its variance accounted for by the capacity. In the second study, a variant of another attention control test, the Stroop task, was applied, which minimized the load on capacity, and no significant contribution of this task to Gf was found. Thus, when contribution of control and capacity were unconfounded, attention control mechanisms of WM contributed to fluid intelligence to a lesser extent than did the mechanisms related to the active buffer of WM.