Generate images with FLUX 3 Image Klein for free, without an account: the open-weights model Black Forest Labs published on 15 October 2026, 5.6 billion parameters, native 4K output, readable text and faithful prompt following, in a size that runs on an 8 GB card. Write a sentence and get the image in the embedded demo below.
The small open-weights member of the FLUX 3 family by Black Forest Labs, released on 15 October 2026 under the FLUX Kommunity License, free for commercial use below five million dollars of yearly revenue. It has 5.6 billion parameters, generates native 4K, edits with up to ten reference images and accepts bounding boxes to place elements, and its int8 build runs on 8 GB of VRAM. A distilled checkpoint renders in a few steps; a base checkpoint is meant for fine-tuning.
FLUX.2 [dev] is six times larger and non-commercial; FLUX 3 Klein keeps most of its prompt adherence and text rendering in a model that fits a consumer card. Qwen-Image 2.1, the other open 2026 model, is a 7-billion-parameter transformer with a stronger hand at editing and Chinese text, under a research-only licence. A same-seed comparison of the two with Stable Diffusion 3.5 is in preparation on this site.
This demo is hosted by a third party (Hugging Face Space) and embedded here; it can be slow, queued or temporarily unavailable when the host is busy. Nothing to install, no account needed.
Read the guide: How diffusion models work, from Stable Diffusion to FLUX