Infinigence AI Open-Sources APXInf for Embodied Edge Inference on Jetson Thor

Infinigence AI has announced the launch of the APXInf edge inference engine in collaboration with Tsinghua University and Shanghai Jiao Tong University. The open-source engine is designed for robotic models that need to close the perceive-decide-act loop locally rather than in the cloud. It supports Jetson-class devices and desktop GPUs, achieving a significant reduction in inference time from about 278 ms to under 26 ms.