In this paper, we present a detailed comparative study of various analytical and numerical methods intended for the evaluation of the escape rate over high-energy barriers (transition rate or, equivalently, switching times) in magnetic systems, using the archetypal application-relevant model of a biaxial macrospin. First, we derive a closed-form analytical expression of the transition rate for such a particle, using the general formalism of Dejardin et al. [Phys. Rev. E 63, 021102 (2001)], and define a parameter which determines whether the system is in the low, intermediate, or high damping regimes. Then we carry out a comprehensive analysis of three numerical algorithms: time-temperature extrapolation method, “energy bounce” methods [S. Wang and P. Visscher, J. Appl. Phys. 99, 08G106 (2006)], and the forward-flux sampling [R. J. Allen et al., Phys. Rev. Lett. 94, 018104 (2005)], which appear to be the most promising candidates for evaluating the transition rate using computer simulations. Based on underlying physical principles and peculiarities of magnetic moment systems, we suggest several optimization possibilities, which strongly improve the performance of these methods for our applications. For energy barriers in the range we compare the switching times, which correspondingly span more than 20 orders of magnitude, obtained with all the above-mentioned analytical and numerical techniques. We show that although for relatively small barriers all methods agree well with each other (and with straightforward Langevin dynamics simulations), for larger barriers the differences become significant, so that only the forward-flux method provides physically reasonable results, giving switching times which exceed the prediction of analytical approaches (interestingly, the ratio is nearly constant for a very broad interval of switching times). The reasons for the corresponding behavior of numerical methods are explained. Finally, we discuss the perspectives of the application of the analyzed numerical techniques to full-scale micromagnetic simulations, where the presence of several contributions to the total system energy makes the situation qualitatively different from that for the macrospin approach.