Perplexity Unveils New AI Dispatcher at Computex 2026

At Computex 2026, Perplexity AI demonstrated a hybrid inference system that automatically distributes tasks between user devices and cloud servers. This technology independently decides in real time which processes to run locally and which to offload to more powerful cloud models. CEO Aravind Srinivas showcased the solution during a joint presentation with Intel CEO Lip-Bu Tan. Ixbt.com reports .

The system's key advantage is that users do not need to manually choose where data is processed. For example, sensitive content like financial documents or medical records is processed locally using Intel Core Ultra Series 3 processors, while complex, resource-intensive calculations are securely routed to cloud models. This approach enhances both data privacy and system performance.

This development is a logical continuation of Perplexity's strategy implemented since early 2026. In February, the company introduced the Computer system, which coordinates 19 models including Claude, Gemini, GPT, and Grok. The new hybrid model determines not only the most suitable model for a task but also where to execute it physically—on the local computer or a server.

The main theme of Computex 2026 was bringing AI directly to user devices. While NVIDIA CEO Jensen Huang unveiled the RTX Spark superchip designed to run large language models on PCs, Intel positioned its new processors as the foundation for hybrid AI scenarios. If Perplexity's technology gains traction, demand for these high-performance processors is expected to rise further.