The code is likely the same whether the DB is in-memory or not. So most likely not. In-memory DBs are already faster that disk DBs, so does it really matter? Having different code paths might make [sqlite3_serialize](1) harder too, and make SQLite _bigger_. Not something I wouldn't myself welcome, but many would be against though. [1]: https://www.sqlite.org/capi3ref.html#sqlite3_serialize