Perplexity Computer adding ability to split tasks between local and cloud models

Perplexity Computer is introducing hybrid agentic inference that splits tasks between local and cloud AI models. The system uses a compact model on-device to detect sensitive data like financial records, health information, and personal files, keeping them local. When more powerful frontier models are needed, tasks are routed to server-based AI. This dynamic switching happens task by task, without requiring user input. Most real-world tasks involve both sensitive and complex components, so the system autonomously coordinates local and cloud processing. Unlike prior tools that require upfront local/cloud selection, Perplexity's approach adapts per task, improving privacy and performance for users handling personal or confidential data.