MoWorld: A Flash World Model

2026-07-07

AI Summary

Existing world models struggle to balance inference efficiency and deployment cost, making them ill-suited for high-frame-rate real-time interaction. This work proposes MoWorld, an end-to-end efficient world model that integrates 3D-native data generation, curriculum-based cross-frame pretraining, diffusion denoising step distillation, and a mixed-precision parallel inference framework. Notably, it achieves real-time interaction at 50 FPS on neural processing units (NPUs) for the first time. Departing from conventional paradigms reliant on large-scale video corpora, MoWorld maintains cinematic visual quality while reducing inference costs to 30%–50% of current models, substantially enhancing practical deployability and environmental adaptability.

Abstract

The future of World Models depends not only on scaling model capability, but also on scaling practicality and inference efficiency. High-frame-rate inference enables responsive perception, planning, and control in real-world autonomous systems. To this end, we present MoWorld, a cost-effective yet high-performance Flash World Model with an end-to-end framework spanning data generation, pre-training, distillation, and efficient inference, enabling up to 50 FPS real-time interaction with cinematic visual quality without the need of high-end GPUs. To enable large-scale real-world deployment, MoWorld jointly optimizes model capability and cost throughout the entire development pipeline. Specifically, unlike existing approaches that primarily rely on large-scale video corpora, MoWorld is built upon a scalable 3D-native data engine accumulated from our large-scale 3D vision and generative modeling pipeline, enabling the efficient construction of geometrically consistent training data across diverse real-world and synthetic environments. Based on this foundation, a curriculum cross-frame pre-training strategy for stable and scalable World Model learning, an efficient denoising-step distillation algorithm to reduce diffusion training cost, and a mixed-precision parallel inference framework for low-cost real-time deployment. MoWorld is the first real-time interactive World Model built on the Neural Processing Unit (NPU) and can achieves up to 50 FPS in such the devices, enabling practical and efficient deployment at scale. Comprehensive evaluations demonstrate that MoWorld achieves leading performance; notably, its average inference cost is only 30\%-50\% of that of existing World Models, providing a practical foundation for large-scale real-world applications of World Models. We also demonstrate diverse applications of MoWorld.

Problem

• World Models
• inference efficiency
• real-time interaction
• computational cost
• scalable deployment

Innovation

• World Model
• 3D-native data engine
• distillation
• real-time inference
• NPU