FLUX 3 Klein is the short name forFLUX 3 Image Klein, from Black Forest Labs. The model card states that the 5.6B parameter label refers to the image transformer, and the model does text-to-image, single-image editing, and multi-reference editing with up to 10 images, plus bounding-box layout and in-image text at up to 4K. The Black Forest Labs blog is dated 2026-10-15.
FLUX 3 Klein: The Unofficial Guide
Last verified:
In 30 seconds
Quick facts
| Fact | Value | Source |
|---|---|---|
| Official name | FLUX 3 Image Klein. Also written FLUX.3 [klein] and FLUX 3 Klein. | |
| Released | 2026-10-15 on the Black Forest Labs blog. The Hugging Face repository was created on 2026-10-13 at 10:58 UTC. This site uses 2026-10-15 as the public date. | |
| Parameters | 5.6B for the image transformer, not the text encoder. The model card says: "The 5.6B parameter label refers to the image transformer". | |
| Text encoder | Qwen3-VL-4B-Instruct (Apache-2.0). | |
| Reference images | Up to 10. | |
| Maximum resolution | 4K. | |
| Checkpoints | Klein: distilled, 16 steps, no guidance, and the checkpoint behind the Black Forest Labs API. Klein Base: undistilled, 50 steps, default CFG 3.5, for fine-tuning and LoRA. The GitHub README lists CFG 4. | |
| Prompt upsampler | Required. The model card says: "Do not use the model without a prompt upsampler." The official upsampler, flux-3-image-klein-pup, is a 9B model fine-tuned from Qwen3.5-9B under Apache-2.0. The BF16 file is 18.8 GB. Any LLM used with the official system prompts is also documented. | |
| License | FLUX Kommunity License v.1.0. |
Klein Base defaults to CFG 3.5 on the official template, the model card, and the Black Forest Labs docs. The GitHub README lists 4.
Start here
Run locally
Diffusers, the reference CLI, and ComfyUI.
ComfyUI
Which files to download, where they go, and the built-in workflows.
VRAM
Official memory figures, and where the published claims disagree.
License
The FLUX Kommunity License in plain English. Not legal advice.
Prompt guide
The upsampler, edit instructions, and bounding boxes.
Coming soon
Compare
What changed from FLUX.2 [klein], using the publisher's numbers only.
Coming soon
Sample images
No sample images yet. Images are added only after they are generated for this site and reviewed, with the prompt, seed, steps, precision, GPU, and date. Each one is labeled “AI-generated with FLUX 3 Image Klein”.
Questions
- How many reference images can FLUX 3 Klein use?
- Up to 10. The model card, the GitHub README, and the Black Forest Labs blog all state that limit.
- What does the 5.6B figure count?
- The model card says: "The 5.6B parameter label refers to the image transformer." That count does not include the text encoder. The text encoder named on the model card is Qwen3-VL-4B-Instruct.
- Is pixglen official?
- No. pixglen is an independent, unofficial resource. It is not affiliated with or endorsed by Black Forest Labs.
Sources
Verified on .