This study addresses the challenges of imbalanced upstream-downstream waste matching and insufficient coordination among stakeholders in urban hazardous waste management. A bi-level optimization model for management systems is developed by integrating multi-objective optimization and game theory. Then, the city of Chengdu, China, is analysed as a case study. An upper-level model is designed to formulate an optimal hazardous waste matching scheme by weighing economic costs against social risks and environmental impact. The lower-level model, meanwhile, uses evolutionary game theory to examine strategy interactions among core stakeholders. The results indicate that the optimization scheme facilitates the local disposal of hazardous waste, resulting in a 36.85% reduction in the total cost of the management system. Notably, the transportation segment achieves the most significant reduction, dropping by about 50% compared with pre-optimization levels. The evolutionary game results show that for a given matching scheme, district-level governments and waste-generating enterprises in 90.91% of Chengdu’s regions adopt the proactive strategy of “supervision and implementation.” The volume of waste generation significantly influences the strategies of waste- generating enterprises, which can be guided by the government through adjustments to subsidy and penalty policies. Compared with subsidies, penalties can more effectively steer the system toward ideal equilibrium. This study presents a “scheme optimization-behaviour guidance” framework for pollution control and carbon reduction in urban hazardous waste. Furthermore, the findings can provide a valuable reference for hazardous waste-management policy in megacities.