We use exact quadratic regularization method to solve large-scale multimodal optimization problems. Quadratic regularization allows us to transform multimodal problems into the maximization of the norm of a vector over a convex set. We will show that such a problem is easier to optimize. For the software implementation of this method, only a local solver is needed. This enables us to solve large-scale multimodal problems efficiently. We have solved almost all large-scale multimodal optimization problems from well-known libraries such as GlobalLib, MinlpLib, and PrincetonLib using this method. We have obtained the best solutions for problems with unknown solutions. To verify the effectiveness of global optimization methods, it is more appropriate to use multimodal test problems of unconstrained optimization with unknown solutions. In this way, the most effective method will lead us to the best solutions. For optimization problems with constraints, it is necessary to take into account the accuracy of satisfying the constraints, which is a more difficult problem. The experimental results demonstrate the high practical effectiveness of the exact quadratic regularization method.