This paper is devoted to direct constrained optimisation of blading systems of large power and small power turbines so as to increase their internal efficiency. The optimisation is carried out using hybrid stochastic-deterministic methods such as a combination of a direct search method of Hooke-Jeeves and simulated annealing or a combination of a bat algorithm and simplex method of Nelder-Mead. Among free shape parameters are blade number and stagger angle, stacking blade line parameters and blade section (profile) parameters. One practical example of efficiency optimisation of turbine blading systems is modification of low load profiles PLK-R2 for high pressure (HP) stages of large power steam turbines. Another optimised geometry is that of an ORC radial-axial cogeneration turbine of 50 kWe. Up to 1% efficiency increase can easily be obtained from optimisation of HP blade profiles, especially by making the rotor blade more aft-loaded and reducing the intensity of endwall flows. Almost 2% efficiency rise was obtained for the optimised 50 kWe ORC turbine due to flow improvement at the suction side of the blade.