From the UN IPCC Fifth Assessment Report (released in full last night), Working Group I Contribution To The IPCC Fifth Assessment Report Climate Change 2013: The Physical Science Basis – Final Draft Underlying Scientific-Technical Basis, Chapter 9, Box 9.2: Climate Models and the Hiatus in Global-Mean Surface Warming of the Past 15 Years, page 31, (bold emphasis added):
(c) Model Response Error
The discrepancy between simulated and observed GMST trends during 1998–2012 could be explained in part by a tendency for some CMIP5 models to simulate stronger warming in response to increases in greenhouse-gas concentration than is consistent with observations… This finding provides evidence that some CMIP5 models show a larger response to greenhouse gases and other anthropogenic factors (dominated by the effects of aerosols) than the real world (medium confidence). As a consequence, it is argued in Chapter 11 that near-term model projections of GMST increase should be scaled down by about 10% (Section 184.108.40.206). This downward scaling is, however, not sufficient to explain the model-mean overestimate of GMST trend over the hiatus period.
Another possible source of model error is the poor representation of water vapour in the upper atmosphere… However, this effect is assessed here to be small, because there was a recovery in stratospheric water vapour after 2005…
In summary, the observed recent warming hiatus, defined as the reduction in GMST trend during 1998–2012 as compared to the trend during 1951–2012, is attributable in roughly equal measure to a cooling contribution from internal variability and a reduced trend in external forcing (expert judgment, medium confidence). The forcing trend reduction is primarily due to a negative forcing trend from both volcanic eruptions and the downward phase of the solar cycle. However, there is low confidence in quantifying the role of forcing trend in causing the hiatus, because of uncertainty in the magnitude of the volcanic forcing trend and low confidence in the aerosol forcing trend.
Almost all CMIP5 historical simulations do not reproduce the observed recent warming hiatus.