This study analyses the nature of weekly inflation response to shocks in the Brazilian economy by adopting a generalized quantile autoregression model in which the autoregressive parameter is allowed to be quantile-dependent. We test for unit root at different conditional quantiles of the response variable, by characterizing its asymmetric dynamics along the business cycle. The method allows us to estimate the magnitude, sign, and the significance of actual shocks that affect Brazilian inflation. We evaluate the robustness of results by adopting a bootstrap procedure. Concerning previous studies, we find evidence of stronger asymmetric persistence in inflationary dynamics in which an inflationary shock below the average dissipates very fast when compared to an inflationary impulse occurring above the average. Location, size, and the sign of a random shock might be essential for inflation adjustment towards long-run equilibrium. The results do not support the full inertia hypothesis.