NVIDIA
NVIDIA Jetson Thor T5000 module
Blackwell-generation robotics module rated at up to 2,070 FP4 TFLOPS with 128 GB of memory.
Overview
Jetson Thor targets physical AI and humanoid robotics, combining a Blackwell architecture GPU with a transformer engine and MIG support with a fourteen-core Arm Neoverse V3AE CPU. The 128 GB memory pool is intended for running large vision-language and policy models on the robot itself.
Typical use cases
- Humanoid and mobile robots
- Multi-sensor autonomy
- On-robot vision-language models
Deployment environment
Robot compute bay or autonomous machine, on a carrier board
Key specifications
- AI performance
- Up to 2,070 TFLOPS (FP4)
- GPU
- NVIDIA Blackwell GPU with transformer engine and MIG support
- CPU
- 14-core Arm Neoverse V3AE
- Memory
- 128 GB LPDDR5X
- Power
- 40–130 W configurable
- Software
- JetPack SDK, CUDA, TensorRT
Specifications taken from current manufacturer documentation and last checked on 2026-09-08.
Source: manufacturer documentation
Specifications are summarised from publicly published manufacturer material and are provided for orientation only. Always confirm current figures directly with the manufacturer before purchasing or designing in.