
Perplexity AI introduces a hybrid local-cloud inference system at Computex 2026. The system autonomously routes AI workloads in real time between user devices and cloud models. It enables local execution of simple queries while offloading complex tasks to frontier cloud models. CEO Aravind Srinivas demonstrated the system with Intel CEO Lip-Bu Tan using a Personal Computer agent. The demo processed confidential deal materials using local models. The system reduces latency and bandwidth usage for users during mid-task operations.
Tap to vote and see what everyone thinks.
Tether Brings AI Memory Compression To Consumer Devices
Summary by ByteBrief