Black Forest Labs has released FLUX.3 Image Klein, a new open-weight SI model available immediately in the Diffusers library. The model unifies text-to-image generation, multi-reference image-to-image editing, and a new layout control feature that allows users to define spatial regions using bounding boxes directly in the prompt.

What Happened

The release includes two open checkpoints in the FLUX.3 Image Klein collection on Hugging Face. The distilled checkpoint generates 1024 × 1024 images in 1.93 seconds on an H100 GPU for text-to-image tasks, according to the company. Image-to-image operations take 3.87 seconds, while layout editing tasks require 4.10 seconds. The base checkpoint runs 50 steps with classifier-free guidance and is intended for fine-tuning.

A key addition is layout control, where users can insert JSON lists of regions with bounding boxes (bbox) into the prompt. This allows the SI model to place specific elements in defined areas of a new image or move existing elements within a reference image by specifying source and target boxes. The model also supports quantized weights in NVFP4 and FP8 formats, which can be loaded using the from_single_file method to reduce memory requirements on smaller GPUs.

Why It Matters

For developers and the SI industry, the release simplifies the workflow for generative SI applications by combining multiple capabilities into a single modular pipeline. The model ships as a Modular Diffusers pipeline, so developers can "pop blocks out or add your own," providing significant architectural flexibility. The ability to control layout via bounding boxes offers more precise compositional control than standard text prompts, which is critical for design and advertising tools.

The availability of quantized variants lowers the hardware barrier for running high-quality SI models locally. The company reports that the NVFP4 variant reduces peak allocated memory to 13.4 GiB after loading, compared to 18.80 GiB for the BF16 variant. This reduction enables deployment on GPUs with lower VRAM, potentially expanding the user base for open-weight SI tools.

The Bottom Line

FLUX.3 Image Klein is now available in Diffusers, offering a unified SI model for image generation and editing with new spatial layout controls. The release supports both high-speed inference on powerful hardware and optimized performance on smaller GPUs via quantization, with specific benchmarks showing significant memory savings.