Two Chinese AI Labs Just Built the Same Model

Two top AI labs in China recently launched new models. Even though these teams worked separately, their latest releases are nearly identical. They both chose the same technical paths to make their software faster and smarter.

Both models use a specific hybrid structure and a training tool called Muon. They also share features like gated residuals and compressed indexers. It is very rare to see two different companies land on the exact same recipe like this.

This move suggests that AI experts are starting to agree on the best way to build. When top labs arrive at the same answer, it usually means they found the most efficient design. This could change how future models are built.