# NVIDIA Jetson Thor The ultimate platform for physical AI and robotics. ## Get Yours Today Purchase your new NVIDIA Jetson Thor through one of our partners. [Shop Now](https://store.nvidia.com/jetson/store/) [Module Specifications](#module-specs) | [Developer Kit Specifications](#dev-kit-specs) | [Video](https://www.youtube.com/watch?v=iYT2haVIgSM) Overview ## A Compact Powerhouse for Advanced AI and Robotics NVIDIA® Jetson Thor™ series modules give you the ultimate platform for physical AI and robotics, delivering up to 2070 FP4 TFLOPS of AI compute and 128 GB of memory with power configurable between 40 W and 130 W. They deliver over 7.5x higher AI compute than [NVIDIA AGX Orin™](https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-orin.md), with 3.5x better energy efficiency. ### NVIDIA Jetson Thor Unlocks Real-Time Reasoning for Physical AI The NVIDIA Blackwell-powered robotics supercomputer delivers 2,070 FP4 teraflops to tackle complex applications including agentic AI, high-speed sensor processing and humanoid robotics tasks. [Learn More](https://blogs.nvidia.com/blog/jetson-thor-physical-ai-edge/) ### Ultimate Platform for Physical AI The NVIDIA Jetson AGX Thor Developer Kit and NVIDIA Jetson T5000 module are generally available, empowering developers everywhere to build the future of physical AI. [Get Started](https://developer.nvidia.com/blog/introducing-nvidia-jetson-thor-the-ultimate-platform-for-physical-ai/) Highlights ## Industry-Leading Performance for Humanoid Robots ### AI Performance 2070 TFLOPS1 ### Memory Bandwidth 273 GB/s ### CPU 14 cores 1. 130 W measured performance Benefits ## Blackwell GPU, Sensor Processing, and Robotic AI Software Stack ### Supercomputer for Humanoids Accelerate generative AI and large transformer models at the edge with the 2070 FP4 TFLOPS Blackwell GPU. ### High-Speed Sensor Processing Ingest high-speed sensor data for real-time performance with 4x 25 GbE networking, a camera offload engine, and a Holoscan Sensor Bridge. ### Robotic AI Software Discover a solution designed for humanoid robotics and physical AI applications, powered by the NVIDIA Isaac™ platform and GR00T foundational models. ### Robust Security Deliver end-to-end safety and security across the compute platform, AI models, and the entire edge-to-cloud pipeline. ## Get Started With NVIDIA Jetson Thor The Jetson Thor AGX Developer Kit delivers unmatched performance and scalability for humanoids and physical AI. This video showcases its key features, specifications, and components, as well as how to power it on and initiate first boot. It also demonstrates some of the latest workflows, including [Isaac GR00T N1](https://developer.nvidia.com/isaac/gr00t), [Video Search and Summarization (VSS)](https://build.nvidia.com/nvidia/video-search-and-summarization), and NVIDIA [Holoscan Sensor Bridge](https://www.nvidia.com/en-us/technologies/holoscan-sensor-bridge.md). [Watch the Video](https://www.youtube.com/watch?v=iYT2haVIgSM) [Read the Technical Brief](https://nvdam.widen.net/s/qlflmll7dl/jetson-thor-technical-brief) Workloads ## Unlock Physical AI With the Power of Jetson Thor #### Humanoid Robotics Supercharge humanoid robot development with NVIDIA Isaac GR00T workflows. Jetson Thor gives you the performance you need for real-time control and multi-sensor processing, while GR00T's comprehensive AI software stack supports a wide range of generative AI models and seamless cloud-to-edge integration. Together, they offer a powerful, integrated solution that accelerates the development and deployment of sophisticated humanoid robots, making them more adaptable, responsive, and capable. [Explore Humanoids](https://www.nvidia.com/en-us/use-cases/humanoid-robots.md) #### Spatial Intelligence Create a visual AI agent with the NVIDIA Video Summarization and Search (VSS) workflow. With its powerful [NVIDIA Blackwell GPU](https://www.nvidia.com/en-us/data-center/technologies/blackwell-architecture.md), Multi-Instance GPU (MIG) technology, and suite of accelerators, Jetson Thor can handle real-time video data streaming and AI inference. This makes it ideal for building AI agents that can perform VSS tasks at the edge. Using advanced vision and language models allows for the creation of robust visual AI agents that can take on complex video workflows. [Create Your Own VSS Workflows](https://build.nvidia.com/nvidia/video-search-and-summarization) #### Multi-Sensor Processing Streamline sensor processing in real time with NVIDIA Holoscan. NVIDIA Holoscan accelerates edge AI development by getting sensor data to the GPU for real-time inference. Build and streamline your end-to-end sensor-processing pipelines with high-performance, user-friendly programming options and production readiness, and deploy seamlessly cloud-to-edge on Jetson Thor. The [NVIDIA Holoscan Sensor Bridge](https://www.nvidia.com/en-us/technologies/holoscan-sensor-bridge.md) is a sensor-over Ethernet technology, designed to enable real-time data streaming and simplify high-speed sensor fusion and actuator integration on NVIDIA edge AI platforms. [Learn More About Holoscan](https://developer.nvidia.com/holoscan-sdk) Virtual Incision #### Generative AI Bring the power of generative AI to the physical world. Jetson Thor delivers unmatched performance and scalability, accelerating low-latency, real-time multi-sensor applications. This makes it ideal for running the latest generative AI models. Jetson Thor modules support a wide range of generative AI models, from VLA (Vision Language Action) models to popular LLMs (Large Language Models) and VLMs (Vision-Language Models), ensuring seamless cloud-to-edge integration. [Learn More](https://blogs.nvidia.com/blog/jetson-generative-ai-edge-oss/) --- Specifications ## NVIDIA Jetson Thor Series | | | | | | --- | --- | --- | --- | | | **Jetson AGX Thor Developer Kit** | **Jetson T5000** | **Jetson T4000** | | AI Performance | 2070 TFLOPS (FP4—Sparse) | | 1200 TFLOPS (FP4—Sparse) | | GPU | 2560-core NVIDIA Blackwell architecture GPU with fifth-gen Tensor Cores Multi-Instance GPU (MIG) with 10 TPCs | | 1536-core NVIDIA Blackwell architecture GPU with fifth-gen Tensor Cores Multi-Instance GPU (MIG) with six TPCs | | GPU Max Frequency | 1.57 GHz | | 1.53 GHz | | CPU | 14-core Arm® Neoverse®-V3AE 64-bit CPU 1 MB L2 cache per core 16 MB shared system L3 cache | | 12-core Arm® Neoverse®-V3AE 64-bit CPU 1 MB L2 cache per core 16 MB shared system L3 cache | | CPU Max Frequency | 2.6 GHz | | | | [Vision Accelerator](https://developer.nvidia.com/embedded/pva) | 1x PVA v3 | | | | Memory | 128 GB 256-bit LPDDR5X 273 GB/s | | 64 GB 256-bit LPDDR5X 273 GB/s | | Storage | 1 TB NVMe M.2 Key M Slot | Supports NVMe through PCIe Supports SSD through USB3.2 | | | Video Encode | 2X NVENCODE | | 1X NVENCODE | | Video Decode | 2X NVENCODE | | 1X NVENCODE | | Camera | HSB camera via QSFP slot USB camera | Up to 20 cameras via HSB Up to 6 cameras through 16x lanes MIPI CSI-2 Up to 32 cameras using Virtual Channels C-PHY 2.1 (10.25 Gbps) D-PHY 2.1 (40 Gbps) | | | PCIe\* | M.2 Key M slot with x4 PCIe Gen5 M.2 Key E slot with x1 PCIe Gen5 | Up to Gen5 (x8 lanes) Root port only—C1 (x1) and C3 (x2) Root Point or Endpoint—C2 (x1), C4 (x8), and C5 (x4) | | | USB\* | 2x USB-A (3.2 Gen2) 2x USB-C (3.1) | xHCI host controller with integrated PHY (up to) 3x USB 3.2 4x USB 2.0 | | | Networking\* | 1x 5GbE RJ45 connector 1x QSFP28 (4x 25GbE) | 4x 25GbE | 3x 25GbE | | Display | 1x HDMI 2.0b 1x DisplayPort 1.4a | 4x shared HDMI2.1 VESA DisplayPort 1.4a—HBR3, MST | | | Other I/O | QSFP connector M.2 Key E expansion slot (WLAN/BT, x1 PCIe, USB2.0, UART, I2C, I2S) M.2 Key M connector (NVMe for storage) PCIe x4 lane, I2C, PCIe x2 lane 2x 13-pin CAN header 2x 6-pin automation header LED JTAG connector (2x 5-pin header) 1x fan connector —12V, PWM, and Tach Audio panel header (2x 5-pin) Microfit power jack RTC backup battery connector 2-pin | 5x I2S/2x audio hub (AHUB), 2x DMIC, 4x UART, 4x CAN, 3x SPI, 13x I2C, 6x PWM outputs | 5x I2S/2x audio hub (AHUB), 2x DMIC, 4x UART, 3x SPI, 13x I2C, 6x PWM outputs\*\* | | Power | 40 W–130 W | 40 W–130 W | 40 W–70 W | | Mechanical | 243.19 mm x 112.40 mm x 56.88 mm Thermal Transfer Plate (TTP) and optional fan or heat sink | 100 mm x 87 mm 699-pin B2B connector Integrated Thermal Transfer Plate (TTP) with heatpipe | | \* Refer to the Software Features section of the latest NVIDIA Jetson Linux Developer Guide for a list of supported features. \*\* Low-speed I/O specification is subject to change. View Full Specs View Fewer Specs ## Get Started ### Get Yours Today Purchase your new NVIDIA Jetson Thor through one of our partners. [Shop Now](https://store.nvidia.com/jetson/store/) #### Explore NVIDIA Jetson NVIDIA Jetson™ is the leading platform for robotics and embedded edge AI applications. Its hardware powers edge AI with energy-efficient, high-performance modules designed for robotics, computer vision, and autonomous systems. [Jetson Hardware](https://www.nvidia.com/en-us/autonomous-machines/embedded-systems.md) [Jetson Software](https://developer.nvidia.com/embedded/develop/software) #### Stay Connected NVIDIA Jetson forums provide a vibrant developer community for sharing knowledge, resolving technical issues, and accelerating innovation with the Jetson platform. [Explore NVIDIA Jetson Forums](https://forums.developer.nvidia.com/c/robotics-edge-computing/jetson-embedded-systems/70) NVIDIA Jetson Thor Series Modules Quick Specs | | | | | --- | --- | --- | | | **Jetson T5000** | **Jetson T4000** | | AI Performance | 2070 TFLOPS (FP4—Sparse) | 1200 TFLOPS (FP4—Sparse) | | GPU | 2560-core NVIDIA Blackwell architecture GPU with fifth-gen Tensor Cores Multi-Instance GPU (MIG) with 10 TPCs | 1536-core NVIDIA Blackwell architecture GPU with fifth-gen Tensor Cores Multi-Instance GPU (MIG) with six TPCs | | CPU | 14-core Arm® Neoverse®-V3AE 64-bit CPU 1 MB L2 cache per core 16 MB Shared System L3 Cache | 12-core Arm® Neoverse®-V3AE 64-bit CPU 64 KB I-Cache, 64 KB D-Cache 1 MB L2 cache per core 16 MB Shared System L3 Cache | | Memory | 128 GB 256-bit LPDDR5X 273 GB/s | 64 GB 256-bit LPDDR5X 273 GB/s | | Networking\* | 4x 25GbE | 3x 25GbE | | Storage | Supports NVMe through PCIe Supports SSD through USB3.2 | | | Mechanical | 100 mm x 87 mm 699 pins | | | Power | 40 W–130 W | 40 W–70 W | \* Refer to the Software Features section of the latest NVIDIA Jetson™ Linux Developer Guide for a list of supported features. [View the Jetson AGX Thor Module Datasheet](https://nvdam.widen.net/s/mdn8tjqrzn/robotics-datasheet-update-jetson-thor-modules-nvidia-us-4767587) NVIDIA Jetson AGX Thor Developer Kit Quick Specs | | | | --- | --- | | AI Performance | 2070 TFLOPS (FP4—sparse) | | Module | Jetson T5000 | | GPU | 2560-core NVIDIA Blackwell architecture GPU with fifth-gen Tensor Cores Multi-Instance GPU (MIG) with 10 TPCs | | CPU | 14-core Arm® Neoverse®-V3AE 64-bit CPU 1 MB L2 Cache per core 16 MB Shared System L3 Cache | | Memory | 128 GB 256-bit LPDDR5X 273 GB/s | | Networking\* | 1x 5GbE RJ45 connector 1x QSFP28 (4x 25GbE) | | I/O | QSFP connector HDMI port DisplayPort 2x USB-A 3.2 | 2x USB-C 3.1 Gigabit Ethernet 2x 13-pin CAN header Microfit power jack | | Storage | 1 TB NVMe M.2 Key M slot | | Power | 40 W–130 W | \* Refer to the Software Features section of the latest NVIDIA Jetson Linux Developer Guide for a list of supported features. [View the Jetson AGX Thor Developer Kit Datasheet](https://nvdam.widen.net/s/gdpfkvkrn2/robotics-and-edge-ai-datasheet-jetson-thor-devkit-nvidia-us-4767806-r1-web) ## Get notified when NVIDIA® Jetson Thor™ is available. Welcome back. Not you? Log Out Welcome back. Not you? Clear form