Black Forest Labs Releases 9B FLUX.2 klein Image Model
The new open-weight model offers a more compact, distilled version of the advanced FLUX architecture for text-to-image and editing tasks.
Black Forest Labs has released FLUX.2 klein, a new 9-billion-parameter open-weight model for generating and editing images. The model introduces a distilled, more compact version of the promising FLUX architecture, aiming to make advanced image synthesis more accessible to the research community.
The release is significant because it brings a more manageable version of FLUX—a newer architecture based on flow matching—to a wider audience. Unlike the diffusion models behind tools like Stable Diffusion, FLUX models are designed for potentially faster inference without sacrificing output quality. By creating a "klein" (German for "small") variant, the lab offers a version that is less demanding on computational resources.
According to the model card on Hugging Face, FLUX.2 klein was trained using knowledge distillation. This technique involves training a smaller "student" model to mimic the behavior of a larger, more powerful "teacher" model, resulting in significant gains in efficiency. The model is capable of both creating images from text prompts and performing image editing.
FLUX.2 klein is available for download with some important restrictions. It is released under a Creative Commons Non-Commercial license (CC BY-NC 4.0), limiting its use to research and personal projects.
Sources
- Visit
black-forest-labs/FLUX.2-klein-9b-kv
Hugging Face
More in Text → Image

Microsoft's Mage-Flow packs image editing into 4B
A compact model handles both text-to-image generation and instruction-based edits at native resolution, under a permissive MIT license.
Boogu-Image-0.1 Brings Unified Multimodal to Open Source
A new Apache-licensed model family folds bilingual text-to-image generation and instruction editing into one system.

NVIDIA distills Qwen-Image for few-step generation
A DMD2-distilled build of Qwen-Image trades sampling steps for speed while keeping the original model's output profile.
0 comments
No comments yet. Be the first to weigh in.