Energy consumption and carbon emissions induced by the production of manufacturing enterprises, especially the iron and steel enterprises (ISEs), are leading to severe global climate change. In this case, exploring the effectiveness of production-oriented carbon reduction projects is of great significance in promoting carbon neutrality and global climate governance. Using a unique micro-level dataset, this paper evaluates ISEs’ carbon emission performance (CEP) during China’s 11th five-year plan period (2006–2010), which is the most important period to remove iron and steel overcapacity. In detail, we estimate the meta-frontier non-radial Malmquist CEP index (MNMCPI) and its decomposition, including the efficiency change (EC) index, the best-practice gap change (BPC) index, as well as the technology gap change (TGC) index. Further, we investigate the effectiveness of production-oriented carbon reduction projects by estimating the influences of energy-saving policies on ISEs’ CEP, taking China’s Top 1000 energy-consuming enterprises program (T1000P) as a quasi-natural experiment. Meanwhile, the changes-in-changes model and the least absolute shrinkage and selection operator inference are applied to conduct a series of robustness checks. In addition, a mediation effect model is applied to ascertain three influence mechanisms: the energy consumption structure effect, the environmental regulation effect, and the foreign direct investment effect. We find a 44.7% increase in the MNMCPI during 2006–2010, showing an improving trend of the CEP. According to the decomposition results of the EC, BPC, and TGC indexes, we discover a significant catch-up effect, a significant innovation effect, as well as a slight technology leadership effect in China’s ISEs. However, the TGC index shows a significant group heterogeneity. On the one hand, private and small enterprises are moving the production frontier upward steadily and increasing speed. On the other hand, the TGC indexes in eastern China and key cities for environmental protection show a fast-climbing trend, indicating an increase in the technology leadership effect. As for the policy effects, the T1000P improves the MNMCPI, EC index, and BPC index but fails to promote the TGC index. Moreover, the positive policy effect shows an evident heterogeneity among different ISEs. Specifically, the T1000P has a more evident positive impact on the CEP in private enterprises, large enterprises, and enterprises in mid-western China and non-key cities for environmental protection than in other enterprises. Finally, we confirm that the T1000P can significantly increase the CEP through the environmental regulation effect.