NVIDIA

NVIDIA JetPack SDK

Official Jetson software stack combining the Linux board support package with CUDA-accelerated AI libraries.

Developer workspace with a single-board computer, sensors and code on screen

Overview

JetPack is the supported software stack for every Jetson module. It bundles Jetson Linux — bootloader, kernel, Ubuntu root filesystem, drivers and over-the-air update tooling — with the Jetson AI stack of CUDA-accelerated libraries, developer tools and samples. JetPack 7 targets robotics and generative AI at the edge.

Typical use cases

  • Jetson application development
  • Device provisioning and updates
  • Robotics software stacks

Deployment environment

Development host plus target Jetson device

Key specifications

Type
Board support package plus AI software stack
Components
Jetson Linux (bootloader, kernel, Ubuntu, drivers, OTA) and the Jetson AI stack
Targets
NVIDIA Jetson modules and developer kits
Current generation
JetPack 7, aimed at robotics and generative AI
Licence
NVIDIA SDK licence terms

Specifications taken from current manufacturer documentation and last checked on 2026-09-08.

Source: manufacturer documentation

Specifications are summarised from publicly published manufacturer material and are provided for orientation only. Always confirm current figures directly with the manufacturer before purchasing or designing in.

Profiler output on a dark developer screen
NVIDIA

NVIDIA TensorRT

Inference compiler and runtime family that optimises models for NVIDIA GPUs, including Jetson.

Type
Inference compiler and runtime ecosystem
Components
TensorRT compiler, TensorRT-LLM, Model Optimizer, TensorRT for RTX, TensorRT Cloud
Optimisations
Graph optimisation, layer fusion, FP16 and INT8 calibration
Deployment
Jetson device, workstation or GPU server
Latency-critical inference
Local LLM serving on GPUs
Jetson deployment optimisation
Code editor showing a model optimisation script
Intel

Intel OpenVINO toolkit

Apache 2.0 licensed toolkit for optimising and deploying inference across Intel CPUs, GPUs and NPUs.

Type
Model optimisation and inference toolkit
Licence
Apache 2.0
Targets
Intel CPUs, integrated and discrete GPUs, and NPUs
Deployment
Any Intel-based edge client, industrial PC or server
AI PC and client inference
Industrial vision on x86
Model conversion and quantisation
Terminal showing an inference session starting
Microsoft

ONNX Runtime

MIT-licensed cross-platform inference accelerator with hardware-specific execution providers.

Type
Cross-platform inference and training accelerator
Licence
MIT
Model sources
PyTorch, TensorFlow and Keras, TensorFlow Lite, scikit-learn and others via ONNX
Deployment
Edge device, client PC or server
Cross-vendor edge deployment
AI PC applications
Server and device parity