Infinigence AI, together with Tsinghua University and SJTU, officially open-sources the embodied edge inference engine APXInf, achieving SOTA performance on Pi 0.5

Infinigence AI, together with Tsinghua University and Shanghai Jiao Tong University, has officially open-sourced the embodied edge inference engine APXInf. The engine achieves SOTA (state-of-the-art) performance on the Pi 0.5 model, aiming to solve the "last mile" problem of deploying embodied AI from algorithms to physical entities.

Embodied AI requires AI models to run in real time on physical entities such as robots, placing extremely high demands on edge computing power and inference latency. The open-sourcing of the APXInf engine provides developers with an efficient edge deployment tool, helping to lower the R&D barriers and computing costs for embodied AI robots.

This open-source release not only advances academic research in edge inference for embodied large models, but also accelerates the industry's push toward large-scale deployment of embodied AI technologies. As related infrastructure improves, the autonomous decision-making and interaction capabilities of robots and smart hardware are expected to improve significantly.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.