NVIDIA refreshed its entry-level robotics computer with twice the inference speed.
The Jetson Orin Nano 2, announced Tuesday, doubles the throughput of the Orin Nano Super in the same compact footprint. It also cuts power draw 40 percent at equal performance.
The module delivers 78 trillion operations per second of AI compute. An 8-core Arm CPU and 8GB of memory round out the package, aimed at robots, delivery drones, and vision systems.
Better Tensor Cores and higher memory bandwidth drive the gains. In a 15-watt mode the board stretches efficiency further for battery-powered machines.
NVIDIA’s Deepu Talla says smaller frontier models now match last year’s giants, which puts real-time reasoning on edge devices within reach. The module runs open models tuned for tight memory budgets, including Cosmos, Nemotron, Gemma 4, and Qwen 3.
Wing plans to evaluate the board for drone perception. Matic Robots will adopt it for on-device conversational AI and semantic home mapping. The Super it replaces launched in December 2024 at $249, making this a hardware step, not a software tweak: 78 TOPS versus 67, and eight cores instead of six.
