🚀

Image Generation - FLUX.2 KLEIN 4B

🛠

Text-to-Image Generation & Multi-Reference Image Editing

Generating and editing high-quality images in under one second, on consumer-grade hardware, in a single unified model

What it does

FLUX.2 [klein] 4B is a next-generation image generation and editing model developed by Black Forest Labs. It is the most accessible and fastest model in the FLUX.2 lineup, designed to bring real-time image creation and editing to a much wider audience. In a single unified model, it handles text-to-image generation, image editing, and multi-reference composition, all at speeds that make interactive, real-time creative workflows possible for the first time in this quality tier.

Problem it solves

  • Speed – Generates images in under one second end-to-end, enabling real-time creative workflows previously impossible with models of this quality level
  • Accessibility – Runs on mainstream consumer GPUs (e.g., NVIDIA RTX 3090/4070) with as little as 13GB of video memory, dramatically lowering the hardware barrier
  • Unified generation & editing – Combines text-to-image generation, image editing, and multi-reference image composition in a single model, eliminating the need for separate specialized tools
  • Production-grade quality at compact size – Delivers image quality that competes with much larger models, making it viable for both local development and large-scale production deployments
  • Edge deployment – Compact enough to deploy in latency-sensitive or resource-constrained environments like edge devices or mobile-adjacent infrastructure

Input/Output

  • Input:
    • Text to Image: Describe the image you want in plain language and FLUX.2 [klein] generates it
    • Image Editing: Provide a reference image and a text instruction describing the desired change
    • Multi-Reference Editing: Provide multiple reference images combined with a text instruction to compose or blend elements across sources
  • Options:
    • diffusion steps (1 ~ 8)
      • More steps = better quality but slower generation (default: 4)
    • random seed
    • output image height & width (512 x 512, 768 x 768, 1024 x 1024)
  • Output: Generated image
    • Photorealistic, high-detail output
image
image

Accuracy & Speed

Speed & Scale

Generation speed
Under 1 second end-to-end
Hardware requirement
~13GB VRAM on NVIDIA RTX 3090 / 4070 and above
Inference steps
As few as 4 steps
Deployment options
Local, edge, or cloud production

Technical Details

Architecture
4B parameter rectified flow transformer
Capabilities
Text-to-image, image editing, multi-reference image composition
Supported frameworks
Hugging Face Diffusers, ComfyUI, BFL API
Safety features
Built-in NSFW and content provenance filters; C2PA watermarking on API outputs

Compliance & Provenance

Provider
Open-source (BFL)
Provider type
Specialized
License
Apache 2.0
EU AI Act risk class
Limited Risk
Art. 50 transparency
Required — outputs are marked. See AI Policy §2.
Region availability
Available globally
Training data summary
Provider model card / training data

For more on how we classify models and mark outputs, see our AI Policy.

Model Source