What Is Linux for Tegra (L4T)?

Linux for Tegra, usually shortened to L4T, is the operating system NVIDIA provides for its Jetson edge AI modules. Tegra is the name of the NVIDIA system-on-chip family inside every Jetson, and L4T is the Linux distribution and board support package built for it. NVIDIA now brands it Jetson Linux, but the L4T name survives in release numbers, file names and forums.

When a Jetson is flashed, L4T is what gives it a working operating system: an Ubuntu-based Linux with the bootloader, kernel and drivers tuned for the Tegra processor. JetPack then adds CUDA, cuDNN, TensorRT and the multimedia libraries on top. Every JetPack release is built on exactly one L4T release.

L4T is more than a Linux distribution. It is the full low-level software package that a Jetson needs to boot and use its hardware.

L4T versions are numbered by major release, such as R35 or R36, with revisions within each. The release fixes the kernel, the Ubuntu base and the modules supported.

A driver or application built for one major release will not simply run on another. Moving from R35 to R36, for example, changes the kernel, the Ubuntu base and the CUDA version, so custom drivers have to be rebuilt and the application retested. The file /etc/nv_tegra_release on the module shows which release it is running.

A Jetson module is flashed from an Ubuntu host computer over USB. The module is put into recovery mode, usually by holding a button or linking a pin on the carrier at power-on, and the flashing tool writes the bootloader, kernel and root filesystem to the module’s eMMC or to an NVMe drive. Developer kits can also boot from a prepared SD card image.

For production, the same tools build a reproducible image: a known L4T release, the carrier board’s device tree and drivers, the application and its dependencies. Flashing every system from that one image is what keeps a fleet consistent.

NVIDIA’s L4T supports its own developer kit carriers out of the box. A third-party carrier board usually routes interfaces differently, so it needs its own device tree, pin configuration and sometimes extra drivers. The carrier vendor supplies these as a board support package layered on L4T, and our explainer on board support packages covers how that works.

This is why the carrier vendor’s BSP release, not only NVIDIA’s, decides which L4T versions a system can run. Our Jetson Orin carrier boards list the JetPack versions their vendors support, and our guide to choosing a Jetson carrier board explains what else to check.

Many teams ship the Ubuntu-based root filesystem with their application installed, which is quick to build and familiar to developers. Others build a minimal image with Yocto, using the community meta-tegra layer, to reduce size, boot time and the number of packages to patch. With JetPack 7.2, NVIDIA added an officially supported Yocto path for Orin and Thor; our article on JetPack 7.2 and Yocto on Jetson Orin covers the change.

Whichever route you take, plan how the image will be updated in the field. Over-the-air update tools for Jetson use A/B root filesystem slots, so a failed update can roll back to the last working image instead of leaving a remote system unable to boot.

We take the L4T base and the carrier vendor’s BSP through to a validated production image, flashed onto every Jetson system as standard. Tell us your application and update plan, and we will build the image to match.