FLUX today announced a strategic partnership with NVIDIA, where both parties will engage in deep technical collaboration in the field of AI image generation. The key highlights of this collaboration include performance optimization, enhanced memory efficiency, and an innovative 3D creation workflow.
In terms of hardware compatibility, the FLUX model has been fully optimized for the newly released GeForce RTX 50 series graphics cards. Its development version, FLUX.1, has achieved significant breakthroughs on the RTX 5090 using NVIDIA Blackwell architecture's FP4 computing technology: it can reach double the computing speed of the RTX 4090 with just 10GB of VRAM, greatly enhancing image generation efficiency.
Comparison of BF16 (left) and FP4 (right) in FLUX.1[dev]
More notably, the two parties jointly launched the "NVIDIA AI Blueprint" workflow, which brings revolutionary changes to 3D creation. Creators can directly complete scene layout, including object placement and lighting setup, in 3D software such as Blender, and then use the FLUX NIM microservice to convert the 3D scene into high-quality images. This innovation greatly simplifies the creation process, making 3D-guided AI image generation more intuitive and efficient.
To ensure widespread application of the technology, FLUX plans to release related products on multiple platforms in early February: the optimized FP4 format model will be available on Hugging Face; the FLUX NIM service will be offered through ComfyUI and ai.nvidia.com; and the one-click installation NVIDIA AI Blueprint workflow will be made available on GitHub.
This collaboration marks a new phase in AI image generation technology, enhancing creation efficiency and opening new avenues for 3D artistic creation, promising a revolutionary working experience for creators worldwide.
Official introduction: https://blackforestlabs.ai/flux-nvidia-blackwell/
