m33393

/

llama-65b-gptq-cuda-4bit-32g-safetensors

like 2

It takes ~42-43GB VRAM to run so no. Try two of them perhaps. But then there might be overhead when doing splits.

Thanks for the quick reply. My friend has a game computer with same 24g 4090. I may borrow the entire PC from him, means I will use two PCs not 1 PC with two cards, then any way to run this model?

Just put the card in your computer... use a riser if you have to