Partially. Open weights is not open source.

In machine learning models the term open source has been largely accepted to mean sharing weights and, if necessary, inference code. You can argue if this is an abuse of the term but everyone does it, and saying someone didn’t deliver if they used it and published weights would probably mean saying the same about mistral, meta, etc.

Yes. So say the same thing about them Open source has a definition and abusing that hurts all of us except the billionaires.

I get the "open source" argument, but what is the issue here?

If you are able to reproduce the thing in its entirety and you're given no restrictions on its use, it seems compatible with the spirit of open sourcing things.

The architecture of the model is open source. Not just the weights. You can run the entire thing locally.