Perplexity Introduces Hybrid Local-Server AI Orchestrator at Computex

Perplexity AI has introduced what it calls the first hybrid local-server inference orchestrator, announced in a company blog post. The system, demonstrated at Computex 2026, automatically determines which AI workloads should execute on a user's device and which should run on cloud-based frontier models.

The new system, named Personal Computer, uses local inference for sensitive data such as health or financial records while routing more complex tasks to remote models. It operates autonomously, deciding on a task-by-task basis where each computation should occur. This approach aims to balance accuracy, privacy, and efficiency without requiring user input.

Perplexity unveiled the technology with Intel during the Computex keynote. The orchestrator runs on Intel’s Core Ultra Series 3 chips and is compatible with other hardware, including Nvidia’s RTX Spark. By distributing workloads between local devices and cloud infrastructure, Perplexity aims to reduce reliance on centralized data centers while keeping confidential information on users’ machines.

The company stated that Personal Computer with local inference will be available in July. The release extends Perplexity’s existing orchestration framework, previously used to manage AI models, to now include compute location as part of the decision process.