FLUX.1 open weights
Black Forest Labs' official inference code for the open FLUX.1 models, with mixed weight licences.
Apache-2.0 code; schnell weights Apache-2.0, dev weights non-commercial unless you license them.
This is Black Forest Labs' reference repository for running the open FLUX.1 text-to-image models yourself. It suits developers who want the original weights and inference code, and who check each model's licence first.
What it is
You clone the repository, create a Python 3.10 environment and run pip install -e ".[all]". The README lists FLUX.1 [schnell] plus [dev], Fill, Canny, Depth, Redux, Kontext and Krea [dev] models on Hugging Face, with an optional TensorRT setup inside NVIDIA's PyTorch container.
Who it's for
It suits developers and researchers comfortable with Python and Hugging Face downloads. It does not suit anyone who wants a point-and-click app, or who needs to use [dev] weights commercially without paying for a licence.
Key features
schnell weights
Released under Apache-2.0 on Hugging Face.
dev model family
Fill, Canny, Depth, Redux, Kontext and Krea variants.
Image editing models
Fill for in and outpainting, Kontext for editing.
Structural conditioning
Canny and Depth models with LoRA versions.
TensorRT option
Optional setup in NVIDIA's PyTorch container.
Usage tracking reference
Sample code reports usage for commercial licensing of dev weights.
Up and running in 3 steps
Our honest verdict
What's great
- The code is Apache-2.0 and the [schnell] weights are Apache-2.0 too.
- One repository covers text-to-image, fill, structural conditioning, image variation and editing models.
- Autoencoder weights are also Apache-2.0, per the README.
- BFL documents a commercial licensing route for the non-commercial [dev] weights.
Where it falls short
- FLUX.1 [dev] and its Fill, Canny, Depth, Redux, Kontext and Krea variants use a non-commercial licence.
- The README gives no VRAM figure, so check each Hugging Face model card before you buy a GPU.
- Install is from source in a Python 3.10 environment; it is a code repository, not an app.
- Commercial use of open [dev] weights expects usage reporting via the reference code or a paid licence.
Pricing & license
The code is free under Apache-2.0 and the schnell weights are free. Commercial use of [dev] weights needs a licence from Black Forest Labs, shown at bfl.ai/pricing/licensing.
Open remote AI-training roles
876 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Can I use FLUX.1 commercially?
FLUX.1 [schnell] weights are Apache-2.0. The [dev] family uses a non-commercial licence, and BFL offers commercial licensing at bfl.ai/pricing/licensing.
What GPU do I need to run FLUX.1?
The README does not state a VRAM figure. Check the Hugging Face model card for the variant you choose.
Is there a UI for these models?
This repository is inference code. Tools such as ComfyUI, InvokeAI and SwarmUI list FLUX support in their own docs.
