In this work, we study the effect of a dual luminescence for CdSe quantum dots (QDs) doped by a manganese impurity. In this effect, one line has fast relaxation and corresponds to light emission from CdSe conduction band, whereas the second is very slow in relaxation with the relaxation time of 0.1−1.0 ms. The second line disappears for quantum dots (QDs) with diameters D ≥ 3.3 nm, and therefore, the luminescence becomes tunable by a QD size. The problem is computationally challenging because of large size QDs and a high degeneracy of the energy states. To overcome this problem, we make four assumptions. These assumptions are the following: (1) a QD optical gap is independent of an Mn impurity for small concentrations; (2) we combine electronic structure calculations for medium size CdSe QDs with the effective mass calculations for medium and large QD sizes and match them in the medium size region, assuming that based on assumption 1, the optical spectrum is independent of Mn even for larger QDs; (3) we cut an MnSe 4 fragment out of the middle of Cd 23 Mn 1 Se 24 , Cd 31 Mn 1 Se 32 , and Cd 81 Mn 1 Se 82 QDs and check whether we can quantitatively explain the mechanism of the second, slow relaxation emission in these experiments using the ab initio SAC−CI multiconfiguration method; and (4) the slow luminescence line is independent of a QD size. We have proved theoretically all these assumptions and found that the critical size of a quantum dot, the size when the second luminescence line disappears, is 3.2 nm, and the slow luminescence energy is 2.3 eV in a tetrahedral ligand field. We also study the case for an Mn impurity placed at a QD surface. Then the symmetry of a fragment is C 3v , and the results of the calculations reveal that the slow luminescence energy is 2.47 eV with the critical size D = 2.7 nm, instead of 3.2 nm for a tetrahedral ligand field. We also predict how this energy depends on the length of an Mn−Se bond. The dependencies appear to be the opposite in these two cases. For the tetrahedral symmetry, the luminescence energy grows with the bond length, whereas for the pyramid, C 3v , symmetry (an Mn is at the surface), it goes down. Furthermore, we study a luminescence energy dependence on a Se−Mn−Se pyramid angle. We find that it is angle independent. These results could be useful for CdSe nanocrystal structures different than Wurtzite.