Speech therapy has been widely used as an essential therapy for compensatory articulation errors in nonsyndromic cleft lip and palate patients. We sought to identify potential biomarkers of nonsyndromic cleft lip and palate children after speech rehabilitation based on resting-state fMRI and graph theory techniques. We scanned 28 nonsyndromic cleft lip and palate and 28 typically developing children for resting-state fMRI on a 3T MRI scanner. Functional networks were constructed, and their topological properties were obtained for assessing between-group differences (two-sample t-tests). Also, language clear degree scale scores were obtained for correlation analysis with the topological features in nonsyndromic cleft lip and palate patients. Significant between-group differences of local properties were detected in brain regions involved in higher-order language and social cognition. There were no significant correlations between topological feature differences and language clear degree scale scores in nonsyndromic cleft lip and palate patients. Graph theory provided valuable insight into the neurobiological mechanisms of speech rehabilitation in nonsyndromic cleft lip and palate patients. The global network features, small-world index, nodal clustering coefficient, and nodal shortest path length may represent potential imaging biomarkers for the estimation of effective speech rehabilitation.