> Kudos to Moonshot for making these weights available for download. Lets not fool ourselves and claim these are "open source" by any understanding of the concept though, there are usage restrictions (even if you download them) and also training data isn't clearly broken down either, nor it it actually using a FOSS license.

They will probably never release the training data because that represents a large part of their competitive moat. The same is true of US companies (Google, OpenAI, Meta etc) none of which has released the full training data for its open models.

They use private datasets that cost a lot to acquire, synthetic datasets and a lot copyrighted material for which they don't have licensing.

What matters most is that, with the necessary hardware, I can download a near frontier model, run it and modify it however I want. The other concerns you mentioned are just noise. And if my company is generating $20 million in revenue or serving 100 million users, it can probably afford a relatively inexpensive commercial license.