A Significant number of reservoirs that highly contribute to the global petroleum production and reserve are considered fractured reservoirs. Generally, to consider a reservoir as a naturally fractured reservoir, the fractures need to form an interconnected network. Dual continuum approaches are commonly used to simplify the modeling process. These models require assigning average properties, such as aperture, permeability, and matrix fracture interaction parameters. In this study, a low permeability fractured reservoir with well-connected fracture networks (Major, Medium, and Minor) was investigated using the Discrete Fracture Model (DFN) provided in the CMG software. Different systematic models in 2 and 3 dimensions were defined for the gas-based enhanced oil recovery processes to investigate the impact of the different fracture network types on the recovery. Dimensionless numbers were used to link the properties of these different networks to the properties of the major fractures network. The increase in the aperture ratios decreased the relative importance of the non-major fracture networks. Moreover, the magnitude of the impact was increased with the increase in the reservoir height for the assisted-gravity drainage mode. However, the increase in the intensity ratio decreased the relative importance of including the non-major fracture networks. The results of the fracture networks indicated that the Foam Assisted Water Alternating Gas (FAWAG) process was more sensitive than the Assisted Water Alternating Gas (WAG) and the continuous gas injection (GI) process. Hence, the magnitude of the impact is process deepening. However, the impact magnitude for excluding these non-major fracture networks depends on other factors such as the properties of reservoir fluids. The homogenization process results indicated that the explicit representation of the homogenized fractures network provides a better capturing of the flow behavior than the pseudo-matrix option for using the standard dual continuum modeling approach. However, the hybrid modeling approach is recommended if the error is significant.