What does mtp ud q4 kxl mean? On which rtx?
I ve got running qwen3.6:35b on rtx 3090, and heard, that it could work on as low as 6gb of vram.
Login to reply
Replies (1)
RTX 3090 as well. “MTP” is multi-token prediction. Details here:
“UD-Q4_K_XL” is unsloth’s 4-bit quantized model.
I want to keep context in VRAM for better performance and 27B quantized 4 bit seems to be the sweet spot on the 24G VRAM of the 3090
Qwen3.6 - How to Run Locally | Unsloth Documentation
Run the new Qwen3.6-27B and 35B-A3B models locally!