NVIDIA unveils Jetson Thor, a powerful AI module for robotics, offering significant advancements in real-time reasoning and sensor processing, according to NVIDIA.
The latest innovation in robotics and AI from NVIDIA, the Jetson Thor, is set to revolutionize real-time reasoning for robotic systems. Designed to meet the demands of millions of robotic developers, Jetson Thor delivers an impressive 2,070 FP4 teraflops, tackling complex applications such as agentic AI and high-speed sensor processing, according to NVIDIA.
Advancements in AI Compute and Memory
NVIDIA has announced that Jetson Thor provides a substantial leap in performance compared to its predecessor, the Jetson Orin. It offers 7.5 times more AI compute, 3.1 times more CPU performance, and double the memory, enabling the processing of high-speed sensor data and visual reasoning at the edge. This advancement is crucial for multimodal AI applications, including humanoid robotics.
Industry Adoption and Applications
Companies like Agility Robotics and Boston Dynamics are already integrating Jetson Thor into their systems. Agility Robotics plans to enhance its humanoid robot, Digit, with Jetson Thor, boosting its real-time perception and decision-making capabilities. Similarly, Boston Dynamics is incorporating Jetson Thor into its humanoid robot, Atlas, to leverage server-level compute capabilities directly on the device.
Jetson Thor’s capabilities extend beyond humanoid robots, accelerating applications such as surgical assistants, delivery robots, and industrial manipulators. The platform supports real-time inference for larger, more complex AI models, making it a versatile tool in the robotics field.
Support for Generative Reasoning Models
Jetson Thor is optimized for generative reasoning models, enabling the next generation of physical AI agents to operate in real-time at the edge. It supports popular generative AI frameworks and reasoning models, including Cosmos Reason, DeepSeek, and Llama, among others. This optimization ensures low latency and high performance in real-world applications.
Research and Development Impact
Research institutions like Stanford University, Carnegie Mellon University, and the University of Zurich are utilizing Jetson Thor to advance perception, planning, and navigation models. At Carnegie Mellon, the Robotics Institute is using Jetson Thor to power autonomous robots for medical triage and search and rescue missions.
With the Jetson AGX Thor developer kit, researchers anticipate improvements in AI model performance and sensor-fusion capabilities, enhancing their ability to experiment with robot fleets and complex environments.
Product Availability and Ecosystem
The Jetson Thor family includes a developer kit and production modules, with the developer kit priced at $3,499. The modules are available starting at $2,999 for orders of 1,000 units. NVIDIA’s Jetson ecosystem supports a variety of application requirements, ensuring quick time-to-market for developers.
As NVIDIA continues to innovate in the field of robotics and AI, Jetson Thor stands out as a powerful tool for developers seeking to push the boundaries of real-time reasoning and AI applications in robotics.
Image source: Shutterstock
