ShengShu Technology Outlines Five-Level Roadmap for General World Models
Event summary
- ShengShu Technology presented a five-level roadmap for General World Models (GWMs) at the 2026 World Robot Conference.
- The roadmap spans from world generation (L1) to world orchestration (L5), with current focus on L1-L3.
- Motubrain, released in April 2026, achieved a score of 96.1 on the RoboTwin 2.0 benchmark.
- ShengShu Technology identifies six key challenges for advancing to L4 and L5, including joint evaluation and learning physical dynamics.
The big picture
ShengShu Technology's roadmap reflects the broader industry shift towards more autonomous and integrated AI systems capable of understanding, predicting, and acting in both digital and physical environments. The company's focus on closed-loop feedback systems and unified architectures aligns with the push for more generalizable and adaptable AI models. The success of this roadmap could position ShengShu Technology as a leader in the development of embodied intelligence.
What we're watching
- Technical Execution
- Whether ShengShu Technology can overcome the six key challenges to advance from L3 to L4 and L5.
- Industry Adoption
- The pace at which General World Models gain traction in real-world applications beyond simulation.
- Competitive Positioning
- How ShengShu Technology's roadmap compares to competitors in the race for general embodied intelligence.
Related topics