Flox Catalog Now Hosting NVIDIA CUDA

NEW YORK - SEPTEMBER 10 - Flox today announced that the NVIDIA CUDA Toolkit, libraries and dependencies are now available as prebuilt packages in the Flox Catalog. This collaboration gives developers using GPU acceleration in the open source ecosystem streamlined access to CUDA.

Previously, Nix and Flox users were required to build CUDA-enabled packages such as NVIDIA TensorRT, PyTorch with CUDA, or OpenCV from source—a process that could take hours, or even days, depending on the complexity of a machine learning stack. Now, with NVIDIA's support, Flox can distribute ready-to-install binaries that allow developers to use CUDA-accelerated software immediately upon download.

"Nix's determinism combined with Flox's developer-first platform makes CUDA toolchains reproducible, auditable, and effortless to share across dev, CI, and cloud. Our binary cache will deliver prebuilt, redistributable CUDA Toolkit stacks in seconds, and Flox users will be able to add CUDA to a new project with a single command," said Ron Efroni, Flox Co-Founder and CEO.

Why It Matters

• Faster AI/ML workflows: Turn days into minutes. Installing CUDA-enabled packages like PyTorch, TensorRT, or XGBoost no longer requires multi-hour builds.
• Broader accessibility: Nix and Flox users can now download and run CUDA-accelerated software immediately.
• Enterprise readiness: The initiative validates Flox's role as a trusted commercial open-source vendor in the AI and cloud-native ecosystems.

The prebuilt CUDA packages are available today through the Flox Catalog. Nix users can also access them by adding Flox's binary cache as a substituter in their configuration.

For more information and to get started, visit https://flox.dev.