Jetson AGX Thor is NVIDIA’s high-end Jetson platform for next-generation robotics and edge AI. NVIDIA materials publish up to 2070 FP4 TFLOPS AI compute for the platform.1
This entry converts the published 2070 FP4 TFLOPS figure to the dense INT8 scale. Using the report precision rule where FP4 throughput is 2x FP8/INT8 throughput, the dense INT8-equivalent value is 1035 TOPS, or 1.035e+15 operations per second.1
Total compute: 1.035e+15 dense INT8 operations per second.
Thor-class edge modules may close part of the gap between embedded systems and datacenter-class AI inference nodes while remaining comparable to the rest of this report on the dense INT8 scale.
NVIDIA. “Jetson AGX Thor.” Accessed 2026. https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-thor/. ↩ ↩2