I heard this many times, but I'm guarded about the conclusions. This whole thing depends on people stating how they think.
When I say "I have an internal voice", some people might misunderstand this phenomenon as something more than it is. For example, one might hear it and think I actually experience an audible narrator.
I have an internal monologue, but I don't experience an audible voice. It's not a sensory thing, but there is a clear "lookup word, form phrase" thing going on internally, in which language plays an essential role. If I wasn't curious about how that works, or interested in the subject, perhaps I wouldn't have noticed that it is there.
Similarly, I know I have some picture based thoughts as well. I can't avoid but have it. Maybe if I was more interested on that, I would have noticed it more proeminently than my language-based stuff.
In other words, I do believe we have varying degrees of mental models, but I just can't grasp how can an experiment could validate someone thinks "completely nonverbally". The brain doesn't even have a single "language area", it's spread all over the place and can be highly plastic depending on the individual. How can we even measure how a person thinks?