Yeah I too am a big supporter. There are already all sorts of open models on Hugging Face. The smaller fined tuned models are often better than the big models at the specific tasks they are trained for.

There’s just a big learning curve at how to start using the open weight models and LoRAs on Hugging Face for example