Carbon nanotubes (CNTs) demonstrate extraordinary properties and show great promise for enhancing out-of-plane properties for traditional polymer/fiber composites and enabling functionality. However, current manufacturing challenges hinder realizing their potential. In this research, both experimental and computational efforts were conducted to investigate the micro/nano-flow behaviors during the manufacturing of CNT-integrated multiscale composites with vacuum-assisted resin transfer molding (VARTM), leading to the development of a novel hierarchical multiscale modeling method. Molecular dynamic (MD) simulation offered a reasonable explanation of CNT dispersion and their motion in a polymer solution. Bi-mode finite-extensible-nonlinear-elastic (FENE) dumbbell simulation was used to analyze the influence of CNT length distribution on the stress tensor and shear-rate-dependent viscosity by solving the stochastic Fokker-Planck equations. Based on the simulated viscosity profile and empirical equations from experiments, a macroscale flow simulation model on the finite element method (FEM) was developed and validated to predict resin flow behavior in the VARTM processing of CNT-enhanced multiscale composites. Optimized process design/control can be carried out based on the multiscale modeling simulation results. The proposed multiscale modeling method provides a comprehensive understanding of micro/nano flow in both atomistic details and mesoscale.