Abstract | A mixture model approach is presented for combining the results of different models or analysis methods into a single probabilistic demand model for seismic assessment. In general, a structure can be represented using models of different type or different number of degrees of freedom, each offering a distinct compromise in computational load versus accuracy; it may also be analysed via methods of different complexity, most notably static versus dynamic nonlinear approaches. Employing the highest fidelity options is theoretically desirable but practically infeasible, at best limiting their use to calibrating or validating lower fidelity approaches. Instead, a large sample of low fidelity results can be selectively combined with sparse results from higher fidelity models or methods to simultaneously capitalize on the frugal nature of the former and the low bias of the latter to deliver fidelity at an acceptable cost. By employing a minimal 5 parameter power-law-based surrogate model we offer two options for forming mixed probabilistic seismic demand models that (i) can combine different models with varying degree of fidelity at different ranges of structural response, or (ii) nonlinear static and dynamic results into a single output suitable for fragility assessment.