The Przewalski's horse (Equus caballus przewalskii) became extinct in the wild in the 1960s. Since 1997, captive-bred horses have been released into the Gobi-B Strictly Protected Area (SPA) in southwestern Mongolia, and successful reproduction in the wild started in 1999. In 2002, the population formed 3 harem groups and 1 bachelor group (total 38) in the wild, and 3 harem groups (24) awaited release in summer 2003 within acclimatization enclosures, totaling 62 individuals. We used the stochastic population simulation model VORTEX to (1) identify key variables and their threshold values in population dynamics, (2) predict extinction risk, and (3) optimize project management and release regime by comparing model parameters with our population data. Maximum age of reproduction, foal mortality, and fecundity rates were key factors in population dynamics, while number of released animals, release interval, and duration of supplementation played lesser roles. The severity level of natural catastrophes had the greatest influence on extinction risk and population size according to VORTEX. Assuming a maximum reproductive age of 16 years, an initial population of >140 horses is necessary to achieve a 95% probability of survival over 100 years under the low-severity level of catastrophes scenario. The corresponding extinction risk for high-severity level of catastrophes is 37%, even for initial population sizes >500. The low natural growth rate of the Przewalski's horse may have been the essential prerequisite for extinction in this remote area of Mongolia. However, uncertainty of results was high and limits the predictive value of the model. Comparisons between model parameters with observed population characteristics over the past 10 years reveal some discrepancies that may require readjustment of the model if present trends continue. While our model currently underestimates reproductive rate and foal survival, adult mortality tends to be higher than estimated in the model. We believe that adult survival can be improved in the wild and that the reintroduction program has a realistic chance of success. We stress the importance of an intensive monitoring program of the Przewalski's horse population and consecutive modeling exercises to reevaluate success of this reintroduction program.