Well, I agree the conflicting narratives make it hard to tell, but as far as I know, the models I usually refer to were conducted by supposedly /non-partisan/ researchers. These three researchers used something like 16 different models, and Trump won in every single one of them, with the majority of them showing him winning by a landslide, with only 3 or 4 showing it to be a close race.
Although looking again, I guess they're strictly /economic/ forecast models, which "tend to ignore election polls and personal characteristics of candidates" according to this article.
https://www.politico.com/story/2019/03/21/trump-economy-election-1230495
Still, what they all basically say is that as long as the economy stays strong, he can't lose.