Qwen 2512 low-step lightning lora.
All examples shown are as-is without upscaling or refining or fixes.
I have best experience with the 8-step BF16. The 8-step FP32 seems to break text quite a bit. Works well if you set strength to 0.8.
I did not make this. Do not ask me questions, I do not know the answers. Go ask at the original source for information.
Description
4-Step FP32
FAQ
Details
Downloads
56
Platform
CivitAI
Platform Status
Deleted
Created
2/11/2026
Updated
4/27/2026
Deleted
3/9/2026
Files
Qwen-Image-2512-Lightning-4steps-V1.0-fp32.safetensors
Mirrors
HuggingFace (7 mirrors)
Qwen-Image-2512-Lightning-4steps-V1.0-fp32.safetensors
Qwen-Image-2512-Lightning-4steps-V1.0-fp32.safetensors
Qwen-Image-2512-Lightning-4steps-V1.0-fp32.safetensors
Qwen-Image-2512-Lightning-4steps-V1.0-fp32.safetensors
Qwen-Image-2512-Lightning-4steps-V1.0-fp32.safetensors
Qwen-Image-2512-Lightning-4steps-V1.0-fp32.safetensors
Qwen-Image-2512-Lightning-4steps-V1.0-fp32.safetensors
CivitAI (2 mirrors)
ModelScope CN (1 mirrors)