Well, we take a distributed approach where I work currently. Some of it is in an Excel or a Google Calc spreadsheet, some in Confluence, the project README.md, random Git comments. Sometimes we annotate the classes in the code with documentation and sometimes even generate the documentation from that. It’s all a little confusing TBH. Some of our models are cross domain and have meaning depending on context.
Postgres schema comments (COMMENT ON COLUMN mytable.mycolumn IS 'spline reticulation coefficient'). Otherwise, I've given up on ever maintaining data model documentation and generate ad hoc graphs from different perspectives in the foreign key network: https://di.nmfay.com/exploring-databases-visually