The paper introduces CaloPointFlow II, an advanced generative model for simulating calorimeter showers as point clouds. This model addresses the computational challenges in particle physics experiments by leveraging machine learning techniques. Key improvements over its predecessor include a novel dequantization technique called CDF-Dequantization and a new normalizing flow architecture named DeepSetFlow. The model is evaluated using the Calorimeter Simulation Challenge (CaloChallenge) datasets II and III, demonstrating superior performance in both low-level and high-level classifications. The results show that CaloPointFlow II accurately captures the spatial structure and energy distributions within calorimeter showers, offering a more efficient and accurate simulation compared to traditional methods. The study highlights the model's potential for enabling more intricate and precise analyses in high-energy physics experiments.The paper introduces CaloPointFlow II, an advanced generative model for simulating calorimeter showers as point clouds. This model addresses the computational challenges in particle physics experiments by leveraging machine learning techniques. Key improvements over its predecessor include a novel dequantization technique called CDF-Dequantization and a new normalizing flow architecture named DeepSetFlow. The model is evaluated using the Calorimeter Simulation Challenge (CaloChallenge) datasets II and III, demonstrating superior performance in both low-level and high-level classifications. The results show that CaloPointFlow II accurately captures the spatial structure and energy distributions within calorimeter showers, offering a more efficient and accurate simulation compared to traditional methods. The study highlights the model's potential for enabling more intricate and precise analyses in high-energy physics experiments.