Black Forest Labs' FLUX.3 Image Klein comes to Diffusers with text-to-image, image editing, and bounding-box layout control

Black Forest Labs' open FLUX.3 Image Klein model is available in Diffusers, supporting text-to-image, multi-reference image-to-image, and layout control through JSON bounding boxes written into prompts. Two checkpoints are listed: a distilled model that generates 1024 × 1024 images in about 2 seconds on an H100 with 16 steps and no CFG, and a base model that uses 50 steps with CFG and is positioned for LoRA fine-tuning.

For builders who need precise object placement or reference-image edits, the reported 0–1000 normalized bounding-box prompt format could reduce reliance on separate control pipelines. The trade-off is choosing the faster distilled checkpoint for inference versus the base model for fine-tuning, while the article only cites H100 speed and does not provide quality or latency comparisons across smaller GPUs.