I never saw a paper published in a journal of (pure) math with more than six authors. Is it a rule?
(see this one)
It turns out that someone put out a bibliometric analysis of publication characteristics in mathematics papers this month, which is very convenient for answering the question!
Richard & Sun (2021). Bibliometric analysis on mathematics, 3 snapshots: 2005, 2010, 2015. arXiv:2102.06831
They found a general increase in the number of authors per paper over time, from 2.1 in 2005 to 2.4 in 2015. (Interestingly, this collaboration was both "internal" and "external" - the number of distinct institutions and distinct countries on a paper also increased steadily). The share of papers with five or more authors was 5.4% in 2015, so presumably slightly higher now. Papers with more authors, or from more places, tended to be more highly cited, which I believe is a common phenomenon across most fields.
They used the Web of Science "research area" classification, which as I understand it will group together a few different Web of Science "categories" (which in turn are inferred from the journal the paper was published in). It will probably thus include more interdisciplinary material than a narrowly defined field of "mathematics" might. They do not break down authorship by category, but it's possible to pull the data and do it yourself.
I ran the numbers for 2020 papers in "mathematics" and in "mathematics, applied", filtered to just "articles". Papers in "mathematics" had an average of 2.24 authors; those in "mathematics, applied" had an average of 2.56. 3.4% of papers in "mathematics" had 5+ authors, versus 5.7% of those in "mathematics, applied".
If we limit it to just those papers published in journals which were only classified as "mathematics" and not as eg "logic / mathematics" or "mathematics / mathematics, applied" (about a third of papers were in journals which were in multiple categories), then we get an average of 2.20 authors, and 3.3% with 5+ authors.
So whole we don't have an explicit classification for "pure mathematics", the non-applied group clearly skews towards a slightly smaller number of authors than applied, and the "just mathematics" group ditto. Papers with five or six authors are not unknown, but they are definitely uncommon - only a few percent of papers. (7 or more authors was around 0.4-0.5%, depending which group you looked at).