NVIDIA

Jetson AGX Thor

Production-grade software for NVIDIA's flagship robotics supercomputer. Deploy Jetson AGX Thor fleets with deterministic Linux, secure OTA, and fleet management — built for physical AI.

Avocado target namejetson-agx-thor

• 2070 TFLOPSFP4 AI Performance
• Blackwell2560-core GPU
• 14-coreArm Neoverse-V3AE CPU
• 128GBLPDDR5X Memory
• PCIe Gen5High-Speed I/O
• 40-130WConfigurable Power

Get Started

Overview

The NVIDIA Jetson AGX Thor is the flagship Jetson module, delivering 2070 FP4 TFLOPS — 7.5x the AI performance of AGX Orin at 3.5x better efficiency. Built on the Blackwell GPU architecture with a 14-core Arm Neoverse-V3AE CPU, it includes 128GB LPDDR5X memory, PCIe Gen5, 16x CSI-2 camera lanes, and 4x25GbE networking. It is the definitive platform for humanoid robotics, autonomous systems, and next-generation physical AI. Paired with Avocado OS and Avocado Connect, Thor gains the production-ready software foundation — enterprise Linux, atomic OTA updates, and fleet management — required for mission-critical deployments at scale.

Avocado OS supports the Jetson AGX Thor on the 2026 release.

Specifications

Specification	Value	Notes
CPU	14-core Arm Neoverse-V3AE
GPU	2560-core Blackwell, 96 Tensor Cores
AI Performance	2070 FP4 TFLOPS / 1035 FP8 TFLOPS	Sparse; 1035 / 517 TFLOPS dense
Memory	128GB LPDDR5X (273 GB/s)
Storage	PCIe Gen5 NVMe	1TB in dev kit
Video Decode	Up to 10x 4Kp60 or 4x 8Kp30	H.265, H.264, AV1, VP9
Video Encode	Up to 6x 4Kp60	H.265, H.264
Camera	16x CSI-2 lanes, 20+ cameras
Networking	QSFP 4x25GbE + RJ45 Multi-GbE
Power	40-130W configurable

Use Cases

Humanoid Robotics

Real-time reasoning for general-purpose robotic platforms, with the headroom to run large vision-language-action models on-device.

Autonomous Machines

Multi-sensor fusion for autonomous vehicles and mobile machines, with deterministic Linux and secure OTA for the field.

Edge AI Servers

Large-model inference (LLMs, VLMs) in industrial and medical settings, managed across a fleet with Avocado Connect.

Getting Started

Init, Install, & Build

Follow the Any Supported Target instructions under Getting Started to begin. This target is jetson-agx-thor, available on the 2026 release feed. The provisioning specifics are below.

Connect the Cables

Provisioning flashes the Jetson over USB while it is powered by USB-C PD and held in Force Recovery Mode. The connectors are on the I/O side of the developer kit — see the NVIDIA Jetson AGX Thor Developer Kit hardware layout for the full reference.

1. Connect the included USB-C power supply to the power USB-C port (5b). The Jetson is powered over USB-C PD during flashing.
2. Connect a USB-C cable from your host to the flashing USB-C port (5a) — the port that supports Force Recovery Mode.
3. Optional: for console access, connect a USB-C cable from your host to the debug port (8), located behind the lid cover.

Enter Force Recovery Mode

The buttons are on the front edge of the developer kit: Power (11), Force Recovery (12), and Reset (13).

1. Press and hold both the Force Recovery button (12) and the Reset button (13) for 3 seconds.
2. Release the Reset button (13) only, keeping the Force Recovery button held.
3. After a further 3 seconds, release the Force Recovery button (12).

You don't need to verify this manually — the Avocado CLI and Avocado Desktop detect whether the Jetson is in Force Recovery Mode and report it when you provision.

Provision

Build the project and flash it to the target:

Host machine

avocado build
avocado provision -r dev --profile tegraflash

Run

Disconnect the flashing USB-C cable from 5a, then power-cycle the device by unplugging and reconnecting the USB-C power supply at 5b. It will boot with the provisioned system. The root user is passwordless in the dev runtime used by this guide. Keep the debug USB-C cable on 8 for console access, or SSH in once the device is on the network.