For 10-year trend segments, 6% (CMIP3) or 8% (CMIP5) of the simulated trends are smaller than or equal to the observed trend over the period 1998-2012 - in agreement with a previous positive consistency test for the period 1998-2009(7). However, for the 15-year trend interval corresponding tothe latest observation period 1998-2012 , only 2% of the 62 CMIP5 and less than 1% of the 189 CMIP3 trend computations are as low as or lower than the observed trend. Applying the standard 5% statisticalcritical value (8), we conclude that the model projections are inconsistent with the recent observed globalwarming over the period 1998- 2012. (note, however, that the standard statistical-test terminology,although widely used, is not strictly appropriate in this case; see supplementary material (9)). The inconsistency increases rapidly with increasing trend length. A continuation of the current observed global warming rate for a period of twenty years or longer would lie outside the ensemble of all model-simulated trends