The seismic hazard assessment is an essential tool for site seismic analysis and in general, for disaster mitigation. The selection of ground motion prediction equations (GMPEs) tends to exert a greater influence on the final results of seismic hazard calculations. Currently, in Peru, two GMPEs exists that are derived from local data; however, these models have important shortcomings due to few records used. This study investigates the applicability of foreign GMPEs derived from regional and global data to the Peruvian subduction-zone. To that end, it was compiled a database of subduction-zone strong-motion records, which consists of 484 ground motion records from 118 subduction-type events with moment magnitudes ranging from 5.0 to 8.4, recorded in Peru and northern Chile, between 1966 and 2015. The average log-likelihood method (LLH) of Scherbaum et al. [1] has been applied to assess the goodness of fit of different GMPEs to the data. Results show that for interface models Youngs et al. [2] and Zhao et al. [3] have the best performance; meanwhile, the Abrahamson et al. [4] model shows a good fit for periods greater than 1.0 s, but it would not be applicable to the northern Peruvian coast. For intraslab events, the Abrahamson et al. [4] model shows the best fit to the data followed by the Zhao et al. [5] model. Finally, weights for each selected GMPE were proposed for B and C site classes [6] in order to produce the best calculation of spectral acceleration.