Source: https://huggingface.co/unsloth/Qwen-Image-2512-GGUF/tree/main
💪Train your own model: https://runpod.io?ref=gased9mt
🍺 Join my discord: https://discord.com/invite/pAz4Bt3rqb
Description
FAQ
Comments (3)
I get low quality output like old phone shot with blurry colored noise and grid. Only qwen 2512.
GGUF Q4_0 25steps CFG4.5
Update. The effect is reduced on Q6 and Q8.
Dumcluck51Jan 7, 2026
I tried GGUF Q5 and got an OOM and I have a RTX 4080 with 16GB.
Having success with Q5_K_M (see below) at 35 to 45 steps, cfg < 1.6, euler, sgm_uniform, ModelSamplingAuraFlow node set from 2 to 5. Tried fewer steps but starts to get grainy. It's running at approximately 8 seconds a step on a Mac Studio M3 Ultra w/ 96gb memory