Estimates of impacts from anthropogenic climate change rely on projections from climate models. Uncertainties in those have often been a limiting factor, in particular on local scales. A new generation of more complex models running scenarios for the upcoming Intergovernmental Panel on Climate Change Fifth Assessment Report (IPCC AR5) is widely, and perhaps naively, expected to provide more detailed and more certain projections. Here we show that projected global temperature change from the new models is remarkably similar to that from those used in IPCC AR4 after accounting for the different underlying scenarios. The spatial patterns of temperature and precipitation change are also very consistent. Interestingly, the local model spread has not changed much despite substantial model development and a massive increase in computational capacity. Part of this model spread is irreducible owing to internal variability in the climate system, yet there is also uncertainty from model differences that can potentially be eliminated. We argue that defining progress in climate modelling in terms of narrowing uncertainties is too limited. Models improve, representing more processes in greater detail. This implies greater confidence in their projections, but convergence may remain slow. The uncertainties should not stop decisions being made.