Inversion methods are used to interpret geophysical data for being multivariate problems. The outcome of these methods justifies random sampling. This sampling causes the parameters to vary independently, leading to the slow convergent of global solution with large uncertainty. The study uses Boltzmann distribution/ sampling instead of random sampling with a hybrid particle swarm optimization and grey wolves optimizer (PSOGWO) technique with both the ability to explore and exploit. The Boltzmann-sampling uses probability that a system will be in a specific state as a function of energy instead of evaluating problems based on the error function. The performance is checked based on random and Boltzmann-sampling over magneto-telluric and direct-current data sets, and the result concludes that Boltzmann-sampling converges the solution to optimum solution in noise-free and noisy data better than random sampling with low uncertainty. Thus, Boltzmann-distribution can converge the solution to true model with reduction in uncertainty.