Access is part of the release
The model is available through BFL’s API, Playground and Hugging Face Spaces. For local use, it supports ComfyUI and Diffusers. LoRA training is available from day one through Ostris AI Toolkit and NVIDIA NeMo AutoModel.
Its weights use the FLUX Kommunity License. Companies with annual revenue below $5 million can use the model commercially for free.
The hardware claim has a boundary
The headline capability is specific: the INT8 version runs on NVIDIA cards with 8 GB of VRAM, and the reported speed is on an RTX 4070. That is a useful point of entry, but it does not tell us how performance changes across other cards, or what image quality the quantized model delivers.
I think the more consequential choice is how many routes Black Forest Labs has opened at launch. A low-cost API sits alongside local tools and hosted access, while the license sets a clear threshold for free commercial use. What the announcement leaves unclear is how the model performs beyond that single hardware example; without those comparisons, the 8 GB figure is an invitation to try it, not a full measure of who can use it well.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X