Lineup → Thor → AGX Thor Developer Kit
Current Generation New: Blackwell

Jetson AGX Thor Developer Kit

Blackwell at the edge. 7.5x the AI compute of AGX Orin, with 3.5x better energy efficiency. Real-time reasoning for robotics and physical AI.

MPN: 945-14070-0080-000 MSRP: $2,799 – $3,499
2,070
FP4 TFLOPS
128 GB
LPDDR5X Memory
130 W
Max Power
14
Neoverse-V3AE Cores
Request Quote → Buy from NVIDIA official Amazon affiliate eBay used + new

What it is

The NVIDIA Jetson AGX Thor Developer Kit is a complete development platform for next-generation physical AI: humanoid robots, autonomous vehicles, surgical assistants, agricultural automation. It pairs NVIDIA's Blackwell GPU architecture with 128 GB of unified memory and a 14-core Arm Neoverse-V3AE CPU, delivering up to 2,070 FP4 teraflops of AI compute in a 130-watt envelope.

Compared to the Jetson AGX Orin it replaces at the top of NVIDIA's edge lineup, Thor delivers roughly 7.5x higher AI compute and 3.5x better energy efficiency. JetPack 7 ships with Thor and is purpose-built for generative AI workloads — large language models, vision-language models, and the robotics-oriented Isaac GR00T stack.

Early adopters include Boston Dynamics, Figure, Caterpillar, Medtronic, Amazon Robotics and Agility Robotics. If you're building anything where a robot needs to perceive, reason, and act in real time, this is the platform.

Specifications

Compute

AI Performance2,070 FP4 TFLOPS
GPU ArchitectureBlackwell
CUDA Cores2,560
Tensor Cores96 (5th gen)

CPU & Memory

CPU14-core Neoverse-V3AE
Memory128 GB LPDDR5X
Bandwidth273 GB/s
Storage1 TB NVMe SSD

I/O

Ethernet4x 25 GbE
PCIeGen 5
USBUSB 3.2 Gen2
DisplayDisplayPort 1.4a

Power & Software

Power Range40 – 130 W
Configurable ProfilesYes
JetPack7.0+
OSUbuntu 24.04 LTS

Ask anything about Thor

Trained on the AGX Thor datasheet, JetPack 7 release notes, and the current Jetson AI Lab tutorials. For pricing, quantity, and availability, request a quote.

Try one of these, or ask your own question.
How does Thor compare to AGX Orin 64GB? What models does it run well at the edge? Power requirements for mobile robotics? Is JetPack 7 backwards-compatible?

Developer resources

Official

NVIDIA Product Page

Datasheet, dev kit overview, ordering links.
Community

Developer Forum

Active troubleshooting and JetPack 7 discussion.
Tutorials

Jetson AI Lab

Run vLLM, Ollama, and Hugging Face models on Thor.
Robotics Stack

NVIDIA Isaac & GR00T

Foundation model for humanoid robotics, deployable on Thor.

Need pricing, availability, or volume terms?

Distributors don't always publicize stock levels or honor MSRP. We talk to you about what you actually need, when you need it, and any backup alternatives that meet your timeline.

Request a Quote