The Jetson Lineup Compared: Orin Nano to AGX Thor
Edge AI has a house name, and it is Jetson. NVIDIA's board line runs from the Orin Nano developer kit on a student desk to the AGX Thor kits aimed at humanoid robots, and this directory carries 13 Jetson products with published TOPS figures. The spread is enormous: about 40 TOPS at the bottom, over 2,000 at the top, a 50x range inside one product family.
The generations matter more than the names. Orin parts pair modest TOPS with tiny power envelopes, 7 to 60 W class, built for cameras, kiosks and drones. The Thor generation jumps the ceiling: the T5000 module publishes 2,070 TOPS with 128 GB of memory, which is data center thinking in an embedded form factor, priced like it at 4,999 dollars for the module in our data.
The Jetson range in the directory
| Jetson product | AI TOPS | Memory | Power | US list price |
|---|---|---|---|---|
| NVIDIA Jetson AGX Thor Developer Kit T5000 128GB | 2070 TOPS | 128 GB | 130 W | $5,499 |
| NVIDIA Jetson T5000 Module 128GB | 2070 TOPS | 128 GB | 130 W | $4,999 |
| NVIDIA Jetson T4000 Module 64GB | 1200 TOPS | 64 GB | 70 W | $2,999 |
| NVIDIA Jetson AGX Orin Developer Kit 64GB | 275 TOPS | 64 GB | 60 W | $3,499 |
| NVIDIA Jetson AGX Orin 64GB Module | 275 TOPS | 64 GB | 60 W | $1,999 |
| NVIDIA Jetson AGX Orin 32GB Module | 200 TOPS | 32 GB | 40 W | $1,799 |
| NVIDIA Jetson Orin NX Super 16GB Module | 157 TOPS | 16 GB | 25 W | $599 |
| NVIDIA Jetson Orin NX 16GB Module | 100 TOPS | 16 GB | 25 W | $999 |
| NVIDIA Jetson Orin NX 8GB Module | 70 TOPS | 8 GB | 20 W | $649 |
| NVIDIA Jetson Orin Nano Super Developer Kit 8GB | 67 TOPS | 8 GB | 25 W | $249 |
| NVIDIA Jetson Orin Nano 8GB Module | 40 TOPS | 8 GB | 15 W | $399 |
| NVIDIA Jetson Xavier NX 16GB Module | 21 TOPS | 16 GB | 20 W | $899 |
TOPS across the line
Choosing inside the line is a power and memory question, not a TOPS question. If the model is a detector or a small VLM on a battery or a robot, the Orin Nano and NX parts are the sweet spot and the developer kits include the carrier board you need to start. If the workload is a real LLM at the edge, memory decides, and the honest options are the 64 GB Orin AGX class and up.
Every board here links to its spec page in the directory, and the full edge range including Coral and Hailo parts is on the AI boards page.