A new method of quantum chemical modeling of large and very large (tens of thousands atoms) nanostructures and biomolecules is proposed. Method uses one-electron model Hamiltonian which approximates Hamiltonian of modern density functional methods (DFT). An acceptable accuracy can be achieved in our method simultaneously with an order of magnitude higher performance, which is typical for having lower accuracy semi-empirical methods like PM3. Graphical processor units (GPUs) will be used for direct calculation of sparse density matrix by fast linear scalable methods.
