Will multi thread indexing lead to smaller index?

I am indexing about 17w sentences, and I use bulk. When I indexed use single thread, it end up about 150M, but when I tried use multi thread, it became 110M, is this possible? I used the count api checked, the total number of document is the same.

Is there any way I can find the different between this two index?

The size of the index can vary depending on how segments have or have not merged. If you are indexing at a higher speed using multiple threads it is possible that the initial segments will be larger and therefore merge differently.

Thanks for your reply.