Some ECMWF parameters are described as being valid at the specified time, the validity time. Sometimes they are referred to as instantaneous parameters, and may have "instantaneous" in their name. Some forecast and all analysed parameters fall into this instantaneous category.
Although they are valid at the specified time, and NOT an average over the model time step, such parameters cannot represent variability on short time scales. This arises because the model values are a grid box average, without variability on spatial scales smaller than the grid. In contrast, high frequency observations at a fixed point will show temporal variability due to advection of small scale spatial variability. Therefore, when comparing with high frequency observations, the latter should be temporally averaged so as to remove the variability that is not represented in the model.
Such instantaneous parameters may, or may not, have been averaged in time. For example, hourly instantaneous values might be averaged over a month, in which case they would be archived in a monthly mean "stream". This would be indicated in other metadata in the GRIB header of the data, as would details of the averaging frequency and period. Even in the case of a monthly mean, these parameters would still be referred to as "instantaneous".
Note, that some other forecast parameters (not instantaneous parameters) are not strictly valid at the validity time, but are in fact valid throughout a processing period. During this period, the parameter could be accumulated, averaged or the minimum/maximum could be calculated, from the values at all the model time steps during the processing period. For more information, see ERA5 terminology: analysis and forecast; time and steps; instantaneous and accumulated and mean rates and min/max parameters