4-Step_V3 Qwen-Image-2512-Turbo-LoRA
UpdateFeb 02, 2026 17:27PublishedFeb 02, 2026 17:08
982
0
1
4-Step_V3 Qwen-Image-2512-Turbo-LoRA - 1
Avatar
User_1vtrwc
Type
LoRA
Basic Model
Qwen-Image
Published time
Feb 02, 2026 17:08
File Signature
b06bab2e3fe28a32cf9424cd240ba2613160265205956dc3da66c92cbd3f375b
Wuli-Qwen-Image-2512-Turbo-LoRA-4steps-V3.0-bf16.safetensors
1.10 GB

Reposted from: https://modelscope.cn/models/Wuli-Art/Qwen-Image-2512-Turbo-LoRA

Qwen-Image-2512-Turbo-LoRA is a 4-step or 8-step ultra-fast LoRA for Qwen Image 2512, trained by the Wuli team. While maintaining output quality comparable to the original model, this LoRA delivers a speedup of over 20x⚡️, with a 2x boost coming from CFG distillation and the rest from a significant reduction in inference steps.

January 8, 2026: Released V3.0, further enhancing image colors and detail performance. This version is now compatible with ComfyUI. January 2, 2026: Released V2.0, improving image colors and details. This version is now compatible with ComfyUI. December 31, 2025: Released V1.0 along with its ComfyUI-compatible version (thanks to @System36). We will soon launch a more powerful V2.0 version.

Gallery

No creation yet