Generally averages of all models are used so the variability between them becomes less obvious. (realclimate.org)
Secondly, the eyeball average of the model outputs does not fall particularly close to the endpoint of the temperature record. (realclimate.org)
True, but that's also true of the models; the comparison I posted — well, tried to post — is between the RSS TMT v4, 0 and a weighted average of the model temperatures in 17 different layers of the atmosphere, several of which lie in the stratosphere. (tamino.wordpress.com)