NVIDIA Open GPU Kernel Modules 610.57.04
NVIDIA's dual MIT/GPLv2-licensed, open-source Linux GPU kernel modules, the default and recommended kernel driver for Turing-generation and newer GeForce, RTX, and datacenter GPUs.
Download TAR.GZ Project website ↗Download Mirrors
| Mirror | Region | Download |
|---|---|---|
| GitHub (Official) Primary | Global | Download |
These are NVIDIA’s actual GPU kernel driver modules released under an open MIT/GPLv2 dual license, not to be confused with the fully proprietary .run driver package. Since the 560 driver series, the open kernel modules have been the default and recommended install for Turing-generation GPUs and newer; older Maxwell/Pascal-era cards still require the fully proprietary driver, since the open modules don’t support them.
What’s Actually Open Here
The kernel-space modules (the parts that talk directly to your kernel and load with elevated privileges) are open source and auditable. The userspace components, OpenGL, Vulkan, and CUDA libraries, remain closed source and are installed separately as part of the regular NVIDIA driver package. This is a meaningful, if partial, shift for NVIDIA, whose kernel drivers were previously a frequent point of friction with mainline kernel development and distro packaging.
Required For Newer Hardware
For NVIDIA’s datacenter-class Grace Hopper and Blackwell platforms, the open kernel modules aren’t just recommended, they’re required, since NVIDIA does not ship a proprietary driver variant for that hardware at all.
Installing
Most users should install this through their distribution’s packaged NVIDIA driver rather than building from source directly, since package managers correctly wire up DKMS/akmod integration for kernel updates:
# Ubuntu
sudo apt install nvidia-driver-580-open
# Fedora
sudo dnf install akmod-nvidia-open
Verify Your Download
Release tags are signed and published on the GitHub releases page.