As one of brain functional imaging modalities, electroencephalography (EEG) based current source reconstruction plays an important role in medical diagnostics and neuroscience. The aim of current source reconstruction is to determine the location and amplitude of the source from the measured EEG of a subject, given a volume conductor model of human head. This inverse problem requires a series of many forward simulations of the possible sources so that the accuracy of the forward solution has a direct impact on the accuracy of the reconstructed source. Our study investigated the impact of the dipole models, which is one of the critical components of the source reconstruction, on EEG forward solution accuracy. We used the quasi-static Maxwell equations for defining the physics of the volume conduction, i.e., a mathematical representation between the dipole source and its resulting electric field at the EEG electrode locations. There are three common forms of the dipoles suitable for the finite element conductor model: Venant, partial integration, and subtraction approaches. To evaluate these dipole models, we used a spherical finite element volume conductor that consisted of four compartments (scalp, skull, CSF and brain). Then we created finite element meshes in three different resolutions (fine, middle and coarse) and put two different meshing constraints (regular and coarsest) on the brain compartment in which the dipole source was placed. The error metrics of RE, RDM, and MAG indicated the accuracy of the forward numerical solution with respect to the analytic solution from a quasi-analytic sphere model. The subtraction approach gives the most accurate forward solution for the fine mesh model with the coarsest brain compartment. The RE of the subtraction method is only 0.4%, while those of the Venant and the partial integration method are 2.7% and 7.2%, respectively. For the regular brain compartment model, the RE of the Venant dipole is 1.4%, while those of the subtraction and the partial integration are 2.8% and 2.2% respectively.