I assess the usefulness of the global output gap in forecasting CPI inflation in Russia through pairwise comparison of domestic and global Hybrid New Keynesian Phillips curve specifications in terms of their Root Mean Square Error and absolute error at each date of out-of-sample forecasts. I estimate a huge number of models formed by all possible combinations of predictors, thereby ensuring the results are robust to the choice of specification. Moreover, I consider various proxies for the global output gap and domestic slack including those from other authors, statistical agencies and those of my own. Additionally, I single out the contribution of each predictor to a model’s forecast accuracy both on the whole period and at each date of out-of-sample forecasts. I find that the models with the global gap perform worse than those without for each measure considered when compared in terms of RMSE. However, in the cross-sections of models’ absolute errors at different dates of out-of-sample forecasts, there are periods when global models outperform the domestic ones. Yet both types of output gaps, domestic and global, worsen forecast accuracy. Instead, such predictors as inflation expectations, real effective exchange rate gap, and capacity utilisation improve it, even in the times of crises, when the errors of all models increase dramatically.