Moore Threads GPU Achieves Top Global Benchmark for 5D World Model Training
Chinese AI company Moore Threads has announced a significant achievement in artificial intelligence, with its domestic GPU technology powering the training of a 5D world model that has achieved a top ranking on the global benchmark. The model, developed by Peking University, utilized Moore Threads' MTT S5000 GPU for its training process. This 5D world model has now secured the first position on the Stanford WorldScore, a key metric for evaluating such AI systems.
At the World Artificial Intelligence Conference (WAIC) 2026, Moore Threads also unveiled a strategic vision for AI development, proposing a framework centered around three distinct AI factories. These proposed factories are designed to address specific stages of the AI lifecycle: one focused on model training, another on token generation, and a third dedicated to agent production. This initiative highlights Moore Threads' ambition to build a comprehensive ecosystem for AI development and deployment.
The successful training of a leading 5D world model on domestic Chinese hardware, like the Moore Threads MTT S5000, signifies advancements in China's indigenous AI capabilities. This achievement, validated by a global benchmark like the Stanford WorldScore, suggests a narrowing gap in high-performance AI computing. The proposed three-AI-factory framework by Moore Threads indicates a strategic move towards vertical integration, aiming to control more of the AI value chain from foundational model training to application-level agent deployment. Such a strategy could enhance national AI competitiveness but also raises questions about market concentration and the long-term implications for global AI collaboration and standards. The focus on specialized factories suggests an effort to optimize efficiency and scale within China's AI ecosystem.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.