A high-resolution version of the finite-volume ¡Èclimate model¡É (fvGCM) jointly developed by NASA/DAO and NCAR/CGD has been used operationally at NASA/DAO for data assimilation and stratospheric re-analysis. The fvGCM has been regarded as a climate model because the physics packages were originally developed and tuned for climate simulations. It was therefore unclear if the fvGCM would do well in medium-range weather predictions. In this study we carried out the so-called ¡Ètransplant experiment¡É in which the forecasts from fvGCM were initialized, not with its own data assimilation system, the fvDAS, but with ¡Èforeign data¡É from the NCEP's AVN model. The forecasts were then compared to the official NCEP AVN forecasts using the ¡Èsame¡É initial conditions. Despite many initial problems/disadvantages (e.g., initial shocks due to different physics, totally incompatible land initial conditions, different vertical/horizontal resolution, and different terrain), it is found that the forecasts from fvGCM using NCEP initial conditions are very competitive, particularly for the winter hemisphere. The summer hemisphere scores are not as good, perhaps because of the physics spin-up problem.