Abstract: Layered heavy oil reservoirs are widely distributed hydrocarbon resources and play a crucial role in fulfilling the global increasing demand for energy. Due to the existence of interlayer heterogeneity, however, the traditional commingled steam injection process has been confronted with the challenges of uneven production and poor performance in the field. In this study, to investigate the improvement effects of a separate steam injection process for the layered heavy oil reservoirs, combining the methods of experiments and numerical simulation, the expansion behavior of the heated chamber and production performance of these two steam injection modes (base case and improved case) are compared and analyzed. First, based on the 2D scaling criteria of steam stimulation experiments and actual properties of a typical layered heavy oil reservoir in China, the experimental parameters are obtained. During experiments, to better simulate the field operation condition, a 2D HTHP (high temperature and high pressure) thermal recovery experimental apparatus equipped with a pressure chamber is proposed. From the experimental observations, the advantages of the separate steam injection mode are illustrated from the expansion behavior of the heated chamber and the production performance characteristics. Thereafter, through a history matching of the experimental results, the laboratory-scale numerical simulation model is developed. Then, from the same-scale numerical simulation model, the steam flooding stage of the base case for the layered heavy oil reservoirs is divided into three phases, and the primary features and critical indices of different phases are obtained. Finally, the effects of reservoir properties and operation parameters on production performance and interlayer divergence are discussed. Experimental results show that the separate steam injection mode achieves uniform heated chamber expansion across layers, and the average proportion of heated chamber is 18% higher than that of the commingled steam injection process. Meanwhile, the improved case increases the final oil recovery factor by around 6%. The simulation results of the developed laboratory-scale numerical simulation model are in good agreement with the experimental observations. For the layered reservoirs with an interlayer permeability contrast of the oil layer reaching 3, it is recommended to adopt the separate steam injection mode. In addition, the optimum cyclic steam injection volume for the reservoir is 6000–7000 m3, and the steam injection rate should be no more than 250 m3/d. This paper contributes to a systematic understanding of steam stimulation performance with different steam injection modes for layered heavy oil reservoirs.