In manufacturing environments with a mix of upgraded and new production lines, balancing manufacturing loads becomes complex due to varying levels of digitalization and multiple evaluation criteria. This paper introduces a leader-follower hierarchical game decision-making model to optimize manufacturing load balancing in high-mix production lines. A bi-agent genetic algorithm (BAGA) is developed, employing a Stackelberg game framework where the leader agent focuses on high-level decisions and the follower agent handles low-level scheduling with engineering constraints. Max-plus algebra is utilized to estimate task completion times for conventional lines. A case study demonstrates the effectiveness of the proposed approach.