Comment #865821 by rolan

Perhaps I did utter opinions as fact

The problem is you uttered factually false claims as if factually true. "It's about 1 K" isn't an opinion.

The number of 1K was from the lecture as I was trying to differentiate between the 'raw' effect of CO2 doubling and that indicated by estimates of climate sensitivity

In the context of your original post what mattered was what real sensitivity results since we live in a world with feedbacks and responses, and not what would happen without them. That 1 K is small was something you wanted to make a big deal of. As far as I am concerned, however clear or unclear you made yourself, you represented 1 K as the temperature rise to be expected of doubling CO2.

I'd like to find out more about the predictive models - I only seem to be able to find conflicting, divergent ones

Models diverge a lot, if they include the good ones and the bad ones. In particular, all models which see CO2's greenhouse effect as a non-issue fail miserably in the most data-rich test of models we have, the reconstruction of past climate conditions. If from now on we limit ourselves only to models which pass this test with flying colours, we see many core points of agreement.
The divergence is on the exact size of the sensitivity. As I said before, we can't be very precise about it - but thanks to faster supercomputers, our models can neglect fewer things and use fewer approximations, and we can thus be more precise than in the past. And even obsolete wide error margins precluded a 1 K figure quite a while ago. Sometimes scientists find out the effect of a process, such as evapotranspiration, by making two models differing only in whether they include it. Whether these models reach low-sensitivity or high-sensitivity conclusions matters little to the question of what is the difference due to the process. Bear this in mind when viewing potholer54's discussion of such an evapotranspiration study.
You wanted examples of successful models. The lectures mentioned one of the better ones, MODTRAN. A less than comprehensive list of them is here. Said list does not, for example, include the Sib2 model used in the aforementioned paper.