Abstract: nformation based optimality criteria were adopted in the optimum design of accelerated life testing (ALT) based on non-parametric model. The optimum design of accelerated life testing based on proportional hazards-proportional odds model was processed for constant stress loading and step stress loading respectively. The optimum problem was established by defining Fisher information matrix and variance-covariance matrix based on log-likelihood function. These optimality criteria could improve the accuracy of model parameter estimates effectively and they avoided the drawbacks of the traditional optimality criteria which minimized the integral of asymptotic variance of a reliability-related function over a specified time period under normal operating condition that might result in different optimum solutions with the varying of the length of the time period. Numerical examples were given to demonstrate these methods.