The extent of spring snow cover and summer sea ice across the Northern Hemisphere high-latitudes has decreased significantly over recent decades, trends that are projected to continue. The loss of snow and ice reduces surface reflectivity, which feeds back on the climate system through the surface albedo feedback (SAF) to enhance warming and promote additional melt. The widespread presence of snow and ice at high-latitudes makes SAF a key driver of regional climate change, while a large spread in simulated SAF (a common feature of previous model generations) makes it an important source of uncertainty in climate model projections. In addition, we know that snow and sea ice behave similarly under climate change as they do in the current seasonal cycle forming the basis for documented emergent constraints. This means that taking steps to reduce the model spread in the observable seasonal cycle SAF should result in better constrained future SAF. However, despite knowledge of the constraint on snow for more than a decade, snow albedo feedback spread did not decline from CMIP3 to CMIP5. Therefore, it is important that we not only track the development of models between generations, but also to highlight the structural and parametric changes behind improved or worsened performance so that we can better understand the processes driving SAF. Here, we quantify SAF in the CMIP6 ensemble and compare with both observations and previous generations. We find that several CMIP6 models have made significant progress in better representing SAF. These improvements can be directly tied to a variety of new or updated parameterizations for snow, vegetation, albedo, and sea ice. However, while most models are showing signs of improvement, some outliers still have unrealistic or worsening SAF, which limits the reduction in ensemble spread.