yeah! but you can choose which model and reasoning efforts your sub agents should use, so you can optimise your token spent!

→ View original post on X — @reach_vb