FLUX 3 Image Klein: open weights, 5.6B parameters
The 5.6B parameter model supports native 4K text-to-image generation, image editing, and multi-reference generation using up to 10 reference images.
- Released as base and step-distilled weights under the FLUX Kommunity License, allowing free commercial use for organizations under $5M in annual revenue.
- INT8 and FP8 builds run on consumer GPUs with 8GB of VRAM, generating 1MP images in under 9 seconds on an RTX 4070 via ComfyUI.
- API access starts at $0.009 per 1K image, with prompt upsampling and pixel restoration features.
- Supports LoRA training on day one through Hugging Face Diffusers, Ostris's AI Toolkit, and NVIDIA NeMo AutoModel.
Developers and creators can run native 4K generation and image editing locally on consumer GPUs with as little as 8GB of VRAM.