---
title: flux2-klein-base-9b-distill-lora
canonical_url: "https://www.modelscope.cn/models/AI-ModelScope/flux2-klein-base-9b-distill-lora"
md_url: "https://www.modelscope.cn/models/AI-ModelScope/flux2-klein-base-9b-distill-lora.md"
repository: AI-ModelScope/flux2-klein-base-9b-distill-lora
last_updated: 2026-06-17
pipeline_tag: image-to-image
tasks:
  - image-to-image
base_model:
  - black-forest-labs/FLUX.2-klein-9B
  - black-forest-labs/FLUX.2-klein-base-9B
base_model_relation: adapter
parameters: 994.1M
tensor_type:
  - F32
  - I64
  - BF16
library_name:
  - lora
  - safetensors
  - pytorch
frameworks:
  - pytorch
supports_inference: img2img
downloads: 272
stars: 2
tags:
  - lora
---

# flux2-klein-base-9b-distill-lora

> flux2-klein-base-9b-distill-lora - AI-ModelScope 在 ModelScope 开源的模型。LoRA for Klein Base 9B extracted from Klein 9B using https://github.com/kijai/ComfyUI-FluxTrainer

AI-ModelScope/flux2-klein-base-9b-distill-lora 是 ModelScope 魔搭社区上的 994.1M 参数image-to-image模型，基于 black-forest-labs/FLUX.2-klein-9B、black-forest-labs/FLUX.2-klein-base-9B 构建，并支持在线推理（img2img）。

- **Repository**: AI-ModelScope/flux2-klein-base-9b-distill-lora
- **Tasks**: image-to-image
- **Parameters**: 994.1M
- **Base model**: black-forest-labs/FLUX.2-klein-9B, black-forest-labs/FLUX.2-klein-base-9B
- **Online inference**: img2img
- **Tags**: lora
- **Downloads**: 272
- **Stars**: 2
- **Last updated**: 2026-06-17

Source: https://www.modelscope.cn/models/AI-ModelScope/flux2-klein-base-9b-distill-lora

---

LoRA for Klein Base 9B extracted from Klein 9B using https://github.com/kijai/ComfyUI-FluxTrainer

For the reasons I cannot explain, it produces static noise at strength > 0.5. And sometimes, you have to go as low as 0.1.

Compared to https://civitai.com/models/2324315/klein-4b9b-base-to-turbo-lora it handles CFG much better, and produces higher frequency details

But on the other hand, https://civitai.com/models/2324315/klein-4b9b-base-to-turbo-lora is much closer to Klein 9B in terms of its output