The multiple demand network (MDN) is conceptualized as the core processing system for multi-tasking. Increasing evidence also provides strong support for the involvement of the MDN in fluid intelligence (gF), that is, the ability to solve new problems. However, the underlying neural mechanisms of declining intelligence in old age are poorly explored, particularly whether maintenance of the functional architecture of the MDN can characterize superior intelligence in successful aging. Here, we used eigenvector centrality (EC) to explore the resting-state functional architecture of the MDN in terms of its communication across the entire brain. We found gF to be negatively associated with age and that the MDN EC competitively mediated age-related decline in gF over the aging lifespan, suggesting that excessive cross-talk from the MDN is deleterious for intelligence. Critically, older individuals with comparable MDN EC as younger individuals exhibited superior gF compared with their age-matched counterparts. Taken together, these data provide support for the maintenance of youth-like functional architecture of the MDN and its implication for superior intelligence in successful aging.