Weighted timed automata are timed automata annotated with costs on locations and transitions. The optimal game-reachability problem for these automata is to find the best-cost strategy of supplying the inputs so as to ensure reachability of a target set within a specified number of iterations. The only known complexity bound for this problem is a doubly-exponential upper bound. We establish a singly-exponential upper bound and show that there exist automata with exponentially many states in a single region with pair-wise distinct optimal strategies.