In this talk we compare two methods of correcting the bias of a General Circulation Model; namely statistical correction performed a posteriori or offline as a function of forecast length, and online correction done within the model integration. The model errors of a low resolution GCM are estimated by the 6-hour forecast residual and averaged over several years. Both online and offline corrections substantially reduce the model bias when applied to independent data. Their performance in correcting the model error is comparable at all lead times, but for lead times longer than 1-day the online corrected forecasts outperform the offline corrected forecasts, having smaller RMS errors and larger anomaly correlations. These results indicate that the online correction reduces not only the growth of the bias but also the nonlinear growth of non-constant (state-dependent and random) forecast errors during the model integration.