Thanks to Adam's on-line bibliography and its statistics we may estimate the number of unique printed texts related to alchemy before 1800 -- which is 2,810 so probably not exceeding 3,000 even with more relaxed criteria of what counts as "alchemical". But has anyone tried to estimate the number of unique works in medieval manuscripts? This would include all that was available to would-be alchemists by, say, 1550 -- so both translations and original works, in Latin and in vernacular languages.
Electronic Thorndike-Kibre (eTK) has records for 33,000 witnesses, so obviously much fewer texts, and of these only a fraction can be on alchemy. Perhaps even fewer of those in Electronic Voigts-Kurtz (eVK2) -- both available at: http://cctr1.umkc.edu/cgi-bin/search
Search on "alchemy" (used as a keyword in MS descriptions) returns 3,315 records but matching unique works would be a problem due to differences in incipts or spelling. There are also many items described as recipes but bearing titles of known alchemical treatises (so perhaps excerpted from those).
Maybe in the printed Thorndike-Kibre or elsewhere there is a "learned guess" on it?