From 2011 to 2013, I did research in geophysics with Professor Shaun Lovejoy at McGill. We looked at predictions from four general circulation models of global temperature fluctuations on different time scales, from hours to centuries.
We found that while last-millennium simulations underestimated temperature variability in control runs, they were more realistic in the industrial period when including external forcings (e.g., volcanic or solar). In pre-industrial scenarios, however, multicentennial variability appeared to be underestimated, possibly due to a lack of certain slow climate processes, like land ice, in the models. If you’re interested in learning more, take a look at our paper: Do GCMs predict the climate… or macroweather?