With GISTEMP's new datasets, I thought I'd update (from here) its comparison with CMIP5 models. (Again, GISS's updates are explained here.)
I'm comparing the total warming for each dataset, calculated as linear_slope*time_interval, as time goes by:
I choose the scenario RCP8.5 for comparison, but it doesn't matter much as all the RCPs use the same historical data up to 2005 and get essentially the same results 'til now. (They aren't exactly the same, I guess because the models are spun up to slightly different equilibrium states before beginning their calculation in 1880).