# Perplexity Launches Local-First Agentic Workflows for Mac New local-first mode enables private, credit-efficient agent execution on Apple silicon hardware. - **Source:** [MacStories](https://brifing.si/source/2d6fb241-7483-4085-bcc6-8301b013a2d2) - **Published:** 2026-10-20 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/61e57ada-aba3-47a6-8338-9dc92ee2fbd3 ## Local-First Agent Architecture Perplexity has introduced a local-first mode for its Computer agent, allowing models to operate entirely on Mac hardware. This architecture functions independently of internet connectivity, ensuring that data remains on the local machine. Users now have the explicit capability to manage data transmission to the cloud only when they provide direct approval. ## Hybrid Compute Integration This local mode functions as a strategic expansion of the Hybrid Compute feature launched last month. Users can now choose between cloud-based frontier models or local execution at the start of their workflow. This flexibility allows for a seamless transition between local processing and cloud-based tasks depending on the specific project requirements. ## Model Specifications and Performance The local environment utilizes Qwen 3.6 or PPLX Pro, which is a post-trained version of the Qwen model. By running these models within a securely sandboxed code and tool-use environment, the system maintains high performance. This setup is specifically optimized for Apple silicon Macs, requiring a minimum of 36GB of memory for operation. ## Privacy and Data Security A primary advantage of the local-first approach is the enhancement of user privacy. Because the agentic workflows execute directly on the Mac, sensitive information does not leave the device unless authorized. This sandboxed execution model provides a robust layer of security for users handling confidential data or proprietary code during their daily tasks. ## Subscription Credit Optimization Beyond privacy, the local-first mode offers significant financial benefits for subscribers. By shifting computational loads to local hardware, users can conserve their subscription credits. This allows Perplexity Pro, Max, and Enterprise users to extend the utility of their current plans while still accessing cloud-based models for complex tasks like web browsing. ## Dynamic Cloud Connectivity While the local mode handles core agentic tasks, cloud-based models remain accessible for specific requirements. Users can trigger cloud connectivity to perform external actions such as browsing the internet or accessing online applications. This hybrid mechanism ensures that the agent remains powerful while maintaining the efficiency and privacy of local-first processing. ## Hardware Requirements and Access The feature is currently available to all Perplexity Pro, Max, and Enterprise subscribers. To ensure optimal performance, the company recommends using Apple silicon Macs equipped with at least 36GB of memory. These hardware specifications are necessary to support the computational demands of running advanced models like Qwen 3.6 locally on the device. ## Strategic Shift in AI This development marks a shift toward balancing powerful cloud-based frontier models with the privacy of local execution. By giving users control over where their data is processed, Perplexity is addressing the growing demand for secure, efficient AI tools. This approach sets a new standard for how agentic workflows can be integrated into professional environments. ## Sources covering this story - https://www.macstories.net/news/perplexity-computers-local-first-approach-to-agents/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._