FLUX.3 Image Klein vs Z-Image: which open image model should you use?
FLUX.3 Image Klein edits images from up to 10 references under a revenue-limited license; Z-Image is text-to-image under Apache 2.0. Compare editing, license and memory.
Verdict. Z-Image is the simpler choice for a product, and FLUX.3 Image Klein is the better tool for editing. Z-Image (6B, Apache 2.0) is text-to-image only: Turbo draws photoreal images with English and Chinese text in 8 steps on a 16 GB GPU, and the base checkpoint suits fine-tuning. FLUX.3 Image Klein (5.6B) generates and edits images from up to 10 references and an INT8 build fits in 8 GB of VRAM, but its community license makes commercial use free only below US$5 million of annual revenue, asks commercial users to review outputs, and the weights are gated on Hugging Face. Neither developer's quality claims are independently verified.
Specifications
| Specifications | FLUX 3 Image Klein | Z-Image |
|---|---|---|
| Developer | Black Forest Labs | Tongyi MAI |
| Released | October 15, 2026 | November 26, 2025 |
| Total parameters | 5.6B (Klein (distilled, 16 steps)) | 6.15B (Z-Image-Turbo) |
| Active parameters | All | All |
| Architecture | Diffusion | Diffusion |
| Context window | – | – |
| Input | Text, Image | Text |
| License | FLUX Kommunity (Conditional) | Apache 2.0 (Permissive) |
| Memory for the smallest variant (4-bit) | 8 GB (Klein INT8 (ComfyUI)) | 16 GB (Z-Image-Turbo) |
Performance claims come from the developer and have not been independently verified.
Pick FLUX 3 Image Klein if
- You need editing, such as changing a photo from an instruction or merging several reference images.
- You want layout-controlled generation, where boxes decide where each element goes.
- You have only 8 GB of VRAM, since the INT8 build fits.
- You are a hobbyist or a company under US$5 million in annual revenue.
Pick Z-Image if
- You want Apache 2.0 with few conditions for a commercial product, at any revenue.
- You need English or Chinese text rendered inside images, such as posters.
- You want fast generation, since Turbo takes 8 steps on a 16 GB GPU.
- You plan to fine-tune from the base checkpoint, or you cannot accept a gated download.