Provided in the present invention is a stereo sound box, including: a box body. Two sets of sound generating units are provided on a surface of the box body, correspondingly for a left sound channel and a right sound channel. Each set of sound generating units includes a low sound unit and at least one high sound unit. The low sound units of the two sets of sound generating units are arranged to mirror each other so as to neutralize the inertia of vibrations. The high sound units are arranged outside each low sound unit. In the described stereo sound box, high-fidelity low sound can be maintained even while the box body is kept light.