WINT4 Quantized Diffusion Models (22 models)
⚠️ 已归档 — 这些模型使用的 WINT4 插件 不再维护。请使用 TINT4(torchao 官方 INT4,LoKr / QuaRot / 跨平台)。
模型清单
| 模型类型 | 模型 | 文件名 |
|---|---|---|
| 10er (LTX2.3 NSFW) | EROS V1.2 | 10er/10EROSV1.2-BF16-quarot-64-int4_fixed_meta.safetensors |
| Boogu (OmniGen2) | Base | boogu/boogu-base32_int4.safetensors |
| Edit | boogu/boogu-edit32_int4.safetensors |
|
| Turbo | boogu/boogu-turbo32_int4.safetensors |
|
| F2K (Flux2 Klein) | Base | f2k/F2K-BASE-quarot-64-int4.safetensors |
| Turbo | f2k/F2K-TURBO-quarot-64-int4.safetensors |
|
| Krea2 | RAW | kr2/KR2-RAW-quarot-64-int4.safetensors |
| Turbo | kr2/KR2-TURBO-quarot-64-int4.safetensors |
|
| Qwen | AIO NSFW v23 | qwen/qwen-AIO-NSFW23-quarot-64-int4.safetensors |
| AIO SFW v19 | qwen/qwen-AIO-SFW19-quarot-64-int4.safetensors |
|
| Edit 2511 | qwen/qwen-EDIT-2511-BF16-quarot-64-int4.safetensors |
|
| Wan | i2v High | wan/wan-i2v-High_int4.safetensors |
| i2v Low | wan/wan-i2v-Low_int4.safetensors |
|
| SCAIL2 | wan/wan21-scail2-FP16-quarot-64-int4.safetensors |
|
| Animate | wan/wananimate_int4.safetensors |
|
| Remix H (FP8) | wan/remix/H-I2V-FP8-wanremix_int4_FP16.safetensors |
|
| Remix L (FP8) | wan/remix/L-I2V-FP8-wanremix_int4_FP16.safetensors |
|
| Remix H (FP16) | wan/remix/wanremixFP16-H_int4.safetensors |
|
| Remix L (FP16) | wan/remix/wanremixFP16-L_int4.safetensors |
|
| Z-Image (Lumina2) | Base | zimage/ZIMAGE-BASE-BF16-quarot-64-int4.safetensors |
| Turbo | zimage/ZIMAGE-TURBO-BF16-quarot-64-int4.safetensors |
参数
全部模型:QuaRot=ON, group_size=64(Boogu 强制 gs=32)
使用
- 下载
.safetensors→ComfyUI/models/diffusion_models/ - 用 WINT4 插件 的 WINT4 Model Loader 加载
- 选择对应模型类型
注意
- 仅 Intel XPU(Arc A770/A750/B580/Flex/Ultra)
- 与 TINT4 的 safetensors 互不兼容
- 不再更新。新用户请用 TINT4
相关
| 资源 | 链接 |
|---|---|
| WINT4 插件 | GitHub |
| WINT8 模型 | lhs101/wint8-quantized-models |
| TINT4 插件 | GitHub |
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support