That contradicts this:

https://www.wsj.com/tech/ai/chinas-xi-touts-open-source-ai-a...

So who even knows.

They don't need to be open weight in the long term; once there's an open-weight Fable-level model with 1M context it'll be pretty much good enough for all coding tasks, no need for new models.