Jetson AGX Orin 64GB Developer Kit

Description

Jetson AGX Orin Developer Kit

The Jetson AGX Orin Developer Kit (64GB) delivers server-grade AI performance at the edge with 275 TOPS of computing power—8x faster than its predecessor. Powered by the NVIDIA Ampere architecture with 2048 CUDA cores and 12-core CPU, this flagship edge AI platform is engineered for demanding generative AI, large language model (LLM) deployment and real-time multi-modal applications. The customized version comes with a pre-flashed SSD (1TB/2TB) containing Ubuntu 22.04 system image, ROS2 environment, AI vision examples and offline LLM cases—ready to run out of the box, is suitable for autonomous machines, intelligent robotics and edge AI innovation.

275 TOPS Flagship AI Performance with 64GB Memory

Powered by NVIDIA Ampere architecture featuring 2048 CUDA cores and a 12-core ARM CPU, the Jetson AGX Orin delivers 275 TOPS AI compute and 64GB of high-bandwidth memory—ideal for running large language models, generative AI, and complex vision transformers locally at the edge with 8x the performance of the previous generation.

Pre-configured software to get started

We provide a ready-to-run solution with SSD (1TB/2TB options) pre-flashed with a customized Ubuntu 22.04 system image including CUDA 12.6, TensorRT 10.7, cuDNN 9.18, OpenCV 4.11, ROS2 environment, AI vision demos (YOLOv11, MediaPipe, DeepStream) and offline LLM application cases—saving days of setup time.

Comprehensive AI & ROS Development Resources

Includes exclusive offline/online large model development materials (updated Feb 2026) covering LLM deployment, text-to-image, image-to-text, visual analysis and autonomous agents. Also features 23+ ROS2 video tutorials and advanced vision guides

Multi-functional application, deployable anytime in real-world scenarios

From autonomous robots and drones to smart cities and medical imaging, the Jetson AGX Orin supports full NVIDIA software stack including Isaac for robotics, DeepStream for video analytics and Omniverse for digital twins. Multiple power modes (15W–60W) allow optimization for performance or efficiency across diverse edge scenarios.