My statement about models is a common and oft cited quip used in science.
"All models are wrong, some models are useful."
AS for predicting "with amazing accuracy", in my view, that assertion is not supported by facts. We can't explain the Little Ice Age, of it's end, or the warming in the early 20th century, and we have at least seven models. Why so many? If one is "amazing", why not use that one?
They are all force fit to past "data". There is no other way to devise a model. You take the data available, find a CORRELATING EQUATION, and hope it predicts. Someone else finds a different CORRELATING EQUATION using a different set of assumptions and it too fit existing data, and you hope that predicts. And there are SEVEN of them, last I counted. Seven different models. And then the IPCC gives a range of SAT of 2°C to 7°C in prediction, which is not what I'd call amazingly accurate. The lower range means things will be difficult, the upper range is effectively catastrophic.
This really is a classic story of misapplication of scientific tools, in my view. I've seen it many times. I've done it a few times myself.