Perplexity’s Computer can now run entirely on your Mac, even offline

This local-first capability means that Perplexity's AI model and the agent directing tasks can run directly on an Apple silicon Mac, leveraging at least 36GB of unified memory for optimal performance. This architecture allows the system to process sensitive user data, like emails and messages, and generate outputs in local applications such as Notes, without these files necessarily leaving the device.

The primary benefit of this local processing is enhanced data privacy, as files only interact with cloud services if the user specifically grants permission. This contrasts with the Hybrid Compute model, which already used local processing for private files but still relied on cloud models for research and reasoning tasks. By moving all core functions on-device, Perplexity aims to address concerns about data exposure for tasks not requiring real-time external information.

Users retain the flexibility to authorize cloud access when tasks demand current web information, browser use, or integration with connected cloud applications. The continued availability of Hybrid Compute within the same app provides a spectrum of operational modes, allowing users to balance privacy and the need for up-to-date external data based on the specific task. This approach offers a practical choice for users managing different types of information.