1 of 1
Story summary
- DeepSeek released a technical paper on Manifold-Constrained Hyper-Connections (mHC) on January 1, 2026, by Liang Wenfeng, founder of DeepSeek.
- The paper presents mHC as a training architecture to improve cost-effectiveness and competitiveness against better-funded US rivals.
- 19 DeepSeek researchers tested mHC on models with 3 billion, 9 billion, and 27 billion parameters, reporting stable training and scalability.
- The development signals a trend of collaboration and openness among Chinese AI companies.
