@sudoingX I feel this, but the window is open to use these larger open models to distill/train smaller ones. Not saying that's something just anyone can do, though.
@tunahorse21 It's still my "quick chat/task" model. Unless another comparable qwen comes along I don't see myself moving away from it. It's not like it gets worse with the (so long as you have web search).
@mtrantalainen@karankendre I think you're right based on what he actually said. But of course you know he'd rather have you not be able to run things locally.
What you'd better hope isn't taken seriously is the idea that RAM should be some kind of centrally-controlled resource.
@Agff_@Da7_Tech I think their endgame is a solidly auto-improving model. Maybe they can't force the quality now, but once good auto-improvement is in place it will eventually catch up.