Complex functions of the human brain rely on dynamic coordination of functionally different neural systems. To understand such brain activity at a network level, here we applied the so-called energy landscape analysis (Fig. 1; [1, 2]) to resting-state fMRI data obtained from healthy younger and older adults. This analysis characterizes large-scale brain dynamics within an energy landscape (Fig. 1(e)) by inferring the maximum entropy model from empirical data (Fig. 1(d)). We used this method to measure the correlation between the efficiency of the neural dynamics and a behavioral index . The efficiency of the neural dynamics was defined as the tendency of transitions between two frequently visited activity patterns. We found that in younger adults, a behavioral score quantifying cognitive functions was predicted by the efficiency of neural dynamics in the so-called cingulo-opercular network (CON), whereas that of older adults was correlated with the efficiency in the so-called default-mode network (DMN). These results demonstrate age-related changes in brain dynamics and suggest the importance of investigating large-scale neural dynamics for better understanding human complex cognitive functions.