i know the fact SegWit transaction have bigger size in byte
nested segwit inputs are a little bigger than old style inputs, but bech32 (i.e. native segwit) inputs are a little smaller
the bech32 outputs are smaller not as inputs, it is because the ScriptPub (of type P2WPKH in comparison to P2PKH) doesn't have the same OPs otherwise the transactions spending them are bigger.
every SegWit transaction has overhead. there is the 2 byte marker/flag (=0001), there is a 1 byte leftover CompactInt indicating size of ScriptSig being zero, and finally the CompactInt added to witness indicating the number of items in it.
unless i've missed something in transactions, the total size seems to be the same for P2WPKH (smaller previous tx but bigger tx spending it)