In a recent paper, Blayo et al. <1994> described the simulation of the North Atlantic general circulation during the period October 1992 to September 1993, as obtained from the assimilation of TOPEX/POSEIDON (T/P) altimeter data in a high-resolution, quasi-geostrophic model of the Atlantic basin between 20 ¿N and 60¿. In line with this paper, the purpose of the present study is to investigate the usefulness of the quasi-geostrophic model for predicting the time evolution of the mesoscale flow field from the perspective of operational oceanography. Forecast experiments have been conducted to address the following three questions: (1) how does the model with assimilation differ from a so-called ''pure forecast'' trajectory, as produced in the absence of altimeter data; (2) can the model evolution controlled with T/P altimetric residuals be considered as a decent description of the actual ocean evolution; (3) is the assimilation useful to predict the future of the system at short and medium ranges? The simulations consist of twin experiments obtained from the assimilation of subsets of the original T/P data set (covering, for instance, only part of the total duration of the reference experiment) or from the assimilation of synthetic altimeter data extracted from the reference simulation. In addition, predictability experiments have been carried out to determine the doubling timescale of small errors in the specification of the model initial conditions. The statistics derived from the simulations indicate that the assimilation of altimetry is efficient in reconstructing the past and the present state of the ocean on an operational basis. The procedure is shown to be able to recover both the surface and the deep flow during the assimilation period. The transport of the Gulf Stream is kept consistent with observational evidences as long as altimeter data are taken into account. However, forecasting mesoscale eddies is realistic only for a couple of weeks in advance. It is found that the averaged doubling time of initial errors is about 20 days, and credit of forecasts over more than 2 months in advance is questionable. As expected, predictions of the future are bounded by the predictability timescales inherent to the physical nature of the system. ¿ American Geophysical Union 1996