DiffuserLite is a lightweight and super-fast diffusion planning framework that significantly increases decision-making frequency by employing a plan refinement process (PRP). PRP generates coarse-to-fine-grained trajectories, reducing the modeling of redundant information and leading to a notable increase in decision-making frequency. DiffuserLite achieves a decision-making frequency of 122Hz, 112.7 times faster than previous mainstream frameworks, and demonstrates state-of-the-art performance on D4RL benchmarks. The framework is simple, flexible, and can be easily integrated into other diffusion planning algorithms. It reduces the complexity of the fitted distribution, decreases the search space, and improves planning efficiency. DiffuserLite is designed to support real-time applications, such as robot control and game AI, by achieving a super-high decision-making frequency. The paper also includes experimental results showing the effectiveness of PRP and the performance of DiffuserLite on various benchmarks.DiffuserLite is a lightweight and super-fast diffusion planning framework that significantly increases decision-making frequency by employing a plan refinement process (PRP). PRP generates coarse-to-fine-grained trajectories, reducing the modeling of redundant information and leading to a notable increase in decision-making frequency. DiffuserLite achieves a decision-making frequency of 122Hz, 112.7 times faster than previous mainstream frameworks, and demonstrates state-of-the-art performance on D4RL benchmarks. The framework is simple, flexible, and can be easily integrated into other diffusion planning algorithms. It reduces the complexity of the fitted distribution, decreases the search space, and improves planning efficiency. DiffuserLite is designed to support real-time applications, such as robot control and game AI, by achieving a super-high decision-making frequency. The paper also includes experimental results showing the effectiveness of PRP and the performance of DiffuserLite on various benchmarks.