# Black Forest Labs adds FLUX 3 Image Klein prompt upsampler to Hugging Face

> A 9‑billion‑parameter model that expands short image requests into detailed prompts is now available on Hugging Face under Apache 2.0.

Oossa · 2026-10-15 · https://oossa.com/en/black-forest-labs-adds-flux-3-image-klein-prompt-upsampler-to-hugging-face

Black Forest Labs released the FLUX 3 Image Klein Prompt Upsampler on Hugging Face. It is a 9 billion‑parameter vision‑language model that turns a brief description, up to ten reference images, and optional bounding boxes into a full prompt for the FLUX 3 Image Klein generator. The model is a fine‑tuned version of Qwen3.5‑9B with the vision tower frozen. It runs by default in front of the image model, but can be disabled with a flag.

## The facts

- Model size: 9 B parameters, 18.8 GB file
- Released on Hugging Face: Wed Oct 14 2026

## Why it matters

Artists and developers can generate richer image prompts automatically, saving time when creating detailed text‑to‑image inputs.

## Sources & references

1. [New model on Hugging Face: black-forest-labs/flux-3-image-klein-pup (image-text-to-text)](https://huggingface.co/black-forest-labs/flux-3-image-klein-pup) – Black Forest Labs, 2026-10-14
2. [New model on Hugging Face: black-forest-labs/flux-3-image-klein-base (image-to-image)](https://huggingface.co/black-forest-labs/flux-3-image-klein-base) – Black Forest Labs, 2026-10-13

Last updated: 2026-10-15