Perplexity Computer adds hybrid local-cloud AI for privacy and power

Perplexity is bringing hybrid agentic inference to its Perplexity Computer platform, enabling automatic task splitting between on-device and cloud AI models. The system uses a compact local model to determine when sensitive data—such as financial records, health information, or personal files—should remain on your machine. For tasks requiring the full capability of frontier models, it seamlessly offloads to the cloud. This eliminates the need for users to choose between local and cloud upfront; the orchestration happens in real-time per task. Perplexity highlights that most real-world work involves a mix of sensitive and computationally heavy elements, making this hybrid approach more practical and efficient. The feature is expected to launch in July 2026.

For professionals handling proprietary or personal data, this hybrid approach offers a significant privacy upgrade without sacrificing AI power. By keeping sensitive workloads local while reserving cloud resources for demanding inference, Perplexity Computer aims to deliver both security and performance. The automatic orchestration also reduces token waste, potentially lowering costs for users running complex agentic workflows. This positions Perplexity Computer as a more enterprise-friendly platform, competing with solutions that force users to trade privacy for capability. With the July release, Perplexity is betting that seamless hybrid inference will become the standard for AI-powered productivity tools.