<p>Graph theory has revolutionized our understanding of the brain’s complex networks, spurring interest in characterizing the topological properties of white matter (WM) functional connectome. However, as this approach is relatively new, the reliability and physiological relevance of WM network measures remain unclear. We aimed to test whether WM functional topology is a reliable and meaningful measure by comparing test-retest reliability of network metrics between WM and gray matter (GM) networks, and assessing the role of WM networks across multiple cognitive domains. We constructed the WM connectome in 580 healthy individuals (421 females; mean age 17.9 ± 0.87 years) by parcellating WM functional images into 128 random regions and computing region-to-region time series Pearson correlations. The test–retest reliability of both global network metrics (clustering coefficient, shortest path length, global efficiency, local efficiency, small-worldness, hierarchy, assortativity, and synchronization) and nodal degree was evaluated. We examined whether these metrics could predict fluid intelligence, vigilance, inhibitory control, working memory, and cognitive flexibility, employing support vector regression. Our findings revealed that WM global properties demonstrate reliability comparable to GM networks, with marginally higher test-retest reliability for small-worldness, clustering coefficient, assortativity, and local efficiency. Regions with the most reliable nodal degrees were concentrated in medial and deep WM structures. Furthermore, both global metrics and nodal degrees significantly predicted cognitive performances. Consequently, WM functional topology is as reliable as GM and meaningfully contributes to cognition, challenging traditional views of WM signals as simply noise and underscoring the importance of investigating WM functional topology in future research.</p>