We describe the construction of a new version of the Europe‐wide E‐OBS temperature (daily minimum, mean and maximum values) and precipitation dataset. This version provides an improved estimation of interpolation uncertainty through the calculation of a 100‐member ensemble of realizations of each daily field. The dataset covers the period back to 1950, and provides gridded fields at a spacing of 0.25° x 0.25° in regular latitude/longitude coordinates. As with the original E‐OBS dataset, the ensemble version is based on the station series collated as part of the ECA&D initiative. Station density varies significantly over the domain, and over time, and a reliable estimation of interpolation uncertainty in the gridded fields is therefore important for users of the dataset. The uncertainty quantified by the ensemble dataset is more realistic than the uncertainty estimates in the original version, although uncertainty is still underestimated in data‐sparse regions. The new dataset is compared against the earlier version of E‐OBS and against regional gridded datasets produced by a selection of National Meteorological Services (NMSs). In terms of both climatological averages and extreme values, the new version of E‐OBS is broadly comparable to the earlier version. Nonetheless, users will notice differences between the two E‐OBS versions, especially for precipitation, which arises from the different gridding method used.