01 / The short version
What happened
Black Forest Labs' FLUX.3 Image Klein is available in Diffusers and supports text-to-image and multi-reference image-to-image generation. It introduces layout control using bounding boxes written into prompts on a 0-1000 scale. The distilled checkpoint generates 1024x1024 images in 16 steps by default. The base checkpoint runs 50 steps with classifier-free guidance and is intended for fine-tuning. The transformer is available as a single file in bf16, fp8r, and nvfp4 formats. Diffusers includes DreamBooth LoRA training scripts for both text-to-image and image-to-image tasks.
See the exact references02 / Key takeaways
What you need to know
- 01
FLUX.3 Image Klein supports text-to-image and multi-reference image-to-image generation.
- 02
Layout control is provided via bounding boxes in prompts using a 0-1000 scale.
- 03
The distilled checkpoint defaults to 16 steps for 1024x1024 images; the base checkpoint uses 50 steps with classifier-free guidance.
Keep in perspective
What to watch for
The source text ends mid-sentence while introducing a section on memory consumption and generation speed, so those details are missing.
Go to the source
Exact references
These are the original pages used for this brief. Publisher claims are not independent evaluations.
01Primary source · Hugging FaceRead the original announcementhttps://huggingface.co/blog/flux-3AI-generated from the linked source. It can miss context; verify consequential details in the original. How the radar works