Perplexity, the AI‑driven search and answer engine, rolled out a new hybrid architecture for its Personal Computer agent on Tuesday. The system, set to debut in July, decides in real time which portions of a user’s request should be processed locally and which should be handed off to cloud‑based models.
A lightweight model runs on the user’s machine, handling routine tasks and safeguarding data such as financial records, health information, and personal files. When a query exceeds the local model’s capabilities, the platform forwards that fragment to a more powerful frontier model in the cloud. The split happens automatically; users do not need to choose between local and remote processing.
The move reflects Perplexity’s push to turn personal devices into decentralized data centers. By offloading only the most demanding work to remote servers, the company expects to cut token usage and lower the expense of cloud computing. It also promises to keep private information out of data‑center pipelines.
Personal Computer, the underlying AI agent, already offers Mac users file editing, application control, and web browsing through Perplexity’s Comet browser. A Windows version is slated for release later this year. The hybrid framework was demonstrated with Intel hardware and is compatible with Nvidia’s RTX Spark platform, suggesting broad hardware support.
Industry observers note that the approach could reshape how AI services balance performance, cost, and privacy. By keeping everyday tasks on the device, Perplexity hopes to free up data‑center capacity for truly high‑end workloads, a strategy that could become a new standard for AI‑enabled consumer software.
This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.