> without completely lobotomizing it

The question in case of quants is: will they lobotomize it beyond the point where it would be better to switch to a smaller model like GPT-OSS 120B that comes prequantized to ~60GB.