Black Forest Labs releases FLUX 3 Image Klein models and Prompt Upsampler as open weights

Black Forest Labs releases FLUX 3 Image Klein, a 5.6B open-weights model for text-to-image generation and image editing, on Hugging Face, repackaged for ComfyUI. It takes a prompt and up to 10 reference images and comes in BF16, int8 and NVFP4 versions; NVFP4 requires a Blackwell GPU. The model is released under the FLUX Kommunity License, which permits non-commercial use and provides a commercial-output exception for Qualifying Users, subject to its terms.

Black Forest Labs also releases FLUX 3 Image Klein Prompt Upsampler, a 9B vision-language model fine-tuned from Qwen3.5-9B, under the Apache License 2.0. It expands short image requests into detailed English prompts, accepts up to 10 reference images and optional bounding boxes, and is on by default in the FLUX 3 inference code.

Design Arena announces on X that FLUX 3 Image Klein is now available on its platform. Its launch post says the model runs on a single consumer GPU and supports native 4K output, but these are the claims of that post, not independently verified figures.