Even today, quantum chemical calculations with experimental accuracy are only feasible for small molecules. This statement is especially true if the considered molecule is far from the equilibrium structure, where the overwhelming majority of quantum chemical models break down. The main purpose of this proposal is to develop new quantum chemical methods that are applicable to at least medium-sized molecules and simultaneously provide results sufficiently close to the experimental data and are capable of describing entire potential energy surfaces. The accuracy goal will be achieved through the reduction of the computational cost of high-precision quantum chemical calculations, which are currently practical for molecules of up to 15 atoms. The cost reduction will be accomplished principally by decreasing the number of numerical parameters to be optimized without sacrificing accuracy. To this end, the negligible parameters will be identified and dropped by adopting the corresponding techniques of computer science. The correct behavior of the models for distorted structures will be ensured by developing new approaches that use a linear combination of functions rather than a single function as a starting point for the description of electronic states. Since the programming work associated with the implementation of the proposed schemes is very complex, the project will rely on the automated programming tools previously developed by the proposer. In addition to the outlined challenging tasks, the proposal aims to implement several more straightforward objectives. In particular, the high-accuracy calculations will be extended to molecular properties that are presently not available. Furthermore, the developed methods will be applied to real-life problems, especially in the field of spectroscopy and atmospheric chemistry.