alt Hacker NewsThey mentioned that they using strong quantization (iirc 3bit) and that the model was degradeted from that. Also, they don't have to use transistors to store the bits.
Replies
amelius • yesterday at 6:41 PM
I think they are talking about the transistors that apply the weights to the inputs.
gpt-oss is fp4 - they're saying they'll next try mid size one, I'm guessing gpt-oss-20b then large one, i'm guessing gpt-oss-120b as their hardware is fp4 friendly