black-forest-labs/flux is the official inference repository for the FLUX.1 models published by Black Forest Labs. It was created on August 1, 2024, and contains minimal inference code written in Python. It runs two paths: generating an image or editing one by supplying a text condition and an image condition. The repository license is Apache-2.0, and it has 26029 GitHub stars.
The role of this repository is to run the FLUX open-weight models (models where you download the published weight files yourself and run them) locally. The README splits its instructions into two methods, a local install and a TensorRT-enabled install, and explains the usage and license of each model, down to the commercial license and how usage is tracked. TensorRT is a library that accelerates inference on the GPU.
Still, the boundary is clear. What is included is the minimal inference code the README describes, and this entry has no evaluation numbers for comparing quality. The only settled value is a momentum score of 2.4 (out of 10); no novelty, technical, or quality scores were given. No new version bundled as a release was found either, so this entry covers the repository itself. Apache-2.0 is the condition covering the repository code, while the model weight license and the commercial usage tracking terms have to be checked separately. The installation and execution methods are laid out in the README, so starting with FLUX locally is easy, but this repository holds no basis for comparing quality against other implementations.