Here is a comparison between my calculation of global temperature data (spherical triangulation) and a sample of CMIP6 models blended so as to to match SST and land data (see here)
The model data are computed as temperature anomalies on the same baseline as the data (1961-1990), so they all align themselves during this 30 year timespan. The real comparison is just on the temperatures trends post 1990.
It seems clear that low sensitivity models are more favoured. This result implies that we can probably expect an average rise in temperature since 1960 of about 1.5C in 2050.