By digging into the code and going a little lower-level I was able to do effectively the same thing while close-enough to the lower-level code that it accepted the maxterms argument (of course, you shouldn't have to do this at all and I would consider it a bug):
sage: from mpmath import hyper

I make no guarantee as to the accuracy of the result or if it's even expected be convergent for the given $a$'s and $b$'s.