My POD Always Goes off

Hi, We moved from serverless to Pod for making sure API returns response much faster. But now GPU goes off. Our Server won't make any API call to POD from 2AM-5AM. I have these questions. 1. When my GPU will go off ? I've chosen on-demand 2. How can I make my pod is always 24/7 up. I tried tmux/services even mock api call within the pod Please help me, This is a huge impact for us.