Who will be funding state of the art local models going forward? AI models are never done or good enough. They will have to be trained on new data and eventually with new model architectures. It will remain an expensive exercise.

I could be wrong because I'm not following this too closely, but the open weights future of both Llama and Qwen looks tenuous to me. Yes, there are others, but I don't understand the business model.