Just use abliterated/uncensored self-hosted models. This is why we need them and the hardware to run them. Yes, the API models which will further close the walled gardens will evolve faster than the open-weights ones, but I'm hopeful about the community as a hole rejecting this status quo.

▲

elephanlemon a day ago | parent [-]

Is there hardware you would suggest that you can actually buy today? Pricing is out of control right now, if you can even find stock.

	▲	ed312 a day ago \| parent [-]
		An RTX4090 or a 128gb macbook can generate a pretty reasonable number of tok/sec. Those are not cheap, but within the realm of serious hobbyist. I really hope the AI bubble deflates a bit so consumers can get more powerful hardware at the low thousand $ range.