Doesn’t surprise me. If you can shell out once for a GPU/NPU and not have to keep paying for tokens, that is an easy sell for a great many. And if the human has to review the output anyway getting output that’s a little worse but a lot cheaper is an attractive offer.
I wonder if the 5090 being discontinued has something to do with that…
Grok/OpenAI/Anthropic/Google/MS/etc can’t make money if everyone is just using local models and despite the republicans pretending to be ‘free market’ they’re all about planned economies (planned economies that benefit only them though)
Doesn’t surprise me. If you can shell out once for a GPU/NPU and not have to keep paying for tokens, that is an easy sell for a great many. And if the human has to review the output anyway getting output that’s a little worse but a lot cheaper is an attractive offer.
I wonder if the 5090 being discontinued has something to do with that…
Grok/OpenAI/Anthropic/Google/MS/etc can’t make money if everyone is just using local models and despite the republicans pretending to be ‘free market’ they’re all about planned economies (planned economies that benefit only them though)