Anyone knows any hacks to reduce GPT4 cost utilisation?

I have built an openAI assistant which is on GPT-4-1106-preview, as I have trained it on some books. Now the cost for that comes to $5/day at 100 threads approx.

If there is anyone here who knows any hacks to reduce this cost or any workaround, then pls DM. My MVP is stuck due to this.