Perplexity, an AI-powered search and reply engine, has a brand new strategy to flip private units into decentralized knowledge facilities.
The corporate mentioned Tuesday that it is including a brand new hybrid local-server system to Private Laptop, its AI agent that may work throughout information, apps and the net. Beginning in July, the system will mechanically resolve which elements of a process ought to run immediately on a consumer’s system and which ought to be despatched to extra highly effective AI fashions within the cloud.
A smaller mannequin operating domestically might deal with delicate knowledge and routine work domestically, corresponding to monetary data, well being data and private information. Extra sophisticated work that requires the capabilities of a bigger AI mannequin might nonetheless be despatched to a server.
As we speak we’re saying that hybrid agentic inference is coming to Perplexity Laptop.
Laptop can cut up duties between an area mannequin operating in your machine and frontier fashions within the cloud. This retains non-public knowledge in your system and maximizes token effectivity.
Coming quickly. pic.twitter.com/6t3PrmI1FX
— Perplexity (@perplexity_ai) June 2, 2026
Perplexity says its system will make that call mechanically, breaking a bigger process into smaller elements and routing each to the suitable place. Customers will not want to decide on between an area mannequin and a cloud-based mannequin earlier than getting began.
Private Laptop is at present out there via Perplexity’s Mac app. It expands the corporate’s current Laptop agent with options together with native file modifying, pc use and shopping via Perplexity’s Comet browser. Perplexity additionally mentioned that Private Laptop is coming to Home windows.
Though the present app is offered on Mac, Perplexity is pitching the underlying know-how as a broader system that may work throughout various kinds of {hardware}. The corporate mentioned it unveiled the system with Intel and that the identical framework runs on different native silicon, together with Nvidia’s RTX Spark platform.
Shifting extra work onto customers’ units might additionally cut back the quantity of high-priced cloud computing required to finish AI duties. Perplexity argues that routine work should not eat the identical knowledge middle assets as a request that genuinely wants one of the crucial succesful AI fashions.
