Thinking about Sea Ice and Model Predictions

My sense is that the global climate models (GCM) have generally done a poor job of projecting loss of sea ice and I think this is important because the change in albedo with loss of sea ice is a positive feedback that could trigger other positive feedback loops (like release of CO2 and CO2e gases from permafrost, ocean floor, etc.)

So, I am searching a little to see if my sense about the GCM sloppiness on sea ice is correct.  Here is what I find:

at Arctic Deeply:

“In September 2007, Arctic sea ice levels reached a dramatic and unexpected new low, tumbling to 4.154 million square kilometers (1.6 million square miles) – roughly 40 percent smaller than what it had been in the 1980s. Summer sea ice melt was far outpacing the models produced by scientists, prompting scientists to join together internationally to produce monthly reports on the anticipated state of the Arctic sea ice based on their individual assessments.”  (emphasis added).

So, the models were not doing a good job in 2007 and the scientists jumped in to improve the models and predictions.  Then I find this article in Carbon Brief from 2014:

Why aren’t climate models better at predicting Arctic sea ice loss?

Looks like they had not gotten a lot better from 2007 to 2014.

On current state of sea ice loss science, I find the following websites that may be accurate and informative:

Sea Ice Prediction Network

Arctic Sea Ice Predictions

But the bottom line still appears to be that the models can only be fairly accurate if they are initiated on a monthly basis, so I think that indicates the complexity and dynamic nature of sea ice loss is just beyond our GCM models absent a monthly reset to correct to observed conditions.  Maybe I have this wrong?