We consider a Bayesian model averaging approach for the purpose of forecasting Swedish consumer price index inflation using a large set of potential indicators, comprising some 80 quarterly time series covering a wide spectrum of Swedish economic activity. The paper demonstrates how to efficiently and systematically evaluate (almost) all possible models that these indicators in combination can give rise to. The results, in terms of out-of-sample performance, suggest that Bayesian model averaging is a useful alternative to other forecasting procedures, in particular recognizing the flexibility by which new information can be incorporated.