Perplexity launches local AI assistant for on-device processing with NVIDIA
Perplexity has introduced Portable Computer, an AI assistant that operates locally on a user’s machine, reducing reliance on cloud servers for tasks involving sensitive data. Developed with NVIDIA, the tool is designed for confidential documents, private code, and business workflows, with optional cloud access for broader research needs.
Editor, Lazyfounder

30 SEC SUMMARY
- Perplexity has launched Portable Computer, an AI assistant that runs locally on a user’s machine.
- The tool is designed to handle confidential documents, private code, and sensitive business data without relying on cloud servers.
- Built with NVIDIA, it uses local processing for tasks like data analysis, file work, and complex workflows, reducing the need for Perplexity credits.
- Cloud access is optional and requires user authorization for tasks needing web information or advanced reasoning.
- Initial support is for NVIDIA DGX Spark, with plans to expand to NVIDIA RTX GPU PCs.
TABLE OF CONTENTS
KEY HIGHLIGHTS
- Perplexity launches Portable Computer, a local-first AI assistant for on-device processing.
- Designed for handling confidential documents, private code, and sensitive business information without cloud dependency.
- Built in collaboration with NVIDIA, using models like PPLX 27B, Qwen 3.8 27B, and NVIDIA Nemotron 3.5 Lightning.
- Local tasks do not consume Perplexity credits, making it suitable for high-volume workflows.
- Supports connectors for Slack, Google Drive, Gmail, and GitHub, with dictation and sandboxed code execution running locally.
- Currently available to Perplexity Pro and Max subscribers on NVIDIA DGX Spark with Linux, Windows support planned.
A local-first approach to AI
Perplexity’s Portable Computer is designed to run AI tasks directly on a user’s device, keeping sensitive data like confidential documents, private code, and financial files offline. According to YourStory, this approach reduces exposure to remote servers, which is critical for industries handling regulated or proprietary information.
The tool can read local files, search documents and code, and execute actions on the device without consuming Perplexity credits. This local processing model could be cost-effective for high-volume tasks, as it avoids reliance on cloud-based usage metrics.
Balancing local and cloud capabilities
While Portable Computer operates locally, it is not entirely disconnected from the cloud. For tasks requiring real-time web information, browser access, or advanced reasoning, the tool can switch to cloud processing—though only with explicit user authorization. YourStory reports that this hybrid model allows users to retain control over sensitive data while still accessing broader capabilities when needed.
Hardware and model support
Portable Computer is initially available on NVIDIA’s DGX Spark, a desktop-sized AI computer. It runs on models like PPLX 27B, Qwen 3.8 27B (a post-trained version of Alibaba Group’s Qwen model), and is expected to support NVIDIA Nemotron 3.5 Lightning, a 30B open model.
DGX Spark combines a 20-core Arm CPU, an NVIDIA GPU, and 128 GB of unified memory. According to YourStory, users can install Portable Computer via a one-click setup in Perplexity’s app. Support for NVIDIA RTX GPU PCs is planned for future releases.
Integration with existing tools
Portable Computer is designed to work with popular applications without sacrificing its local-first approach. It supports connectors for Slack, Google Drive, Gmail, and GitHub, allowing users to compare data across platforms while keeping processing on-device. For example, it can cross-reference GitHub issues with Gmail bug reports and organize findings locally.
Dictation features also run locally using NVIDIA Nemotron 3.5 ASR, ensuring spoken notes or private briefings do not leave the device. Perplexity states that the system uses isolated sandbox environments for code and tool execution, limiting AI access to specific tasks and files.
Availability and platform support
The tool is currently available to Perplexity Pro and Max subscribers using NVIDIA DGX Spark, with Linux as the initial supported operating system. Windows support is planned for future updates. YourStory notes that this launch reflects a broader trend: as local hardware becomes more powerful, advanced AI tasks may increasingly shift from cloud-only to hybrid or on-device processing.
What this means
Lazyfounder analysis — our interpretation, not reported fact.
Perplexity’s move underscores a growing demand for AI tools that prioritize data privacy and control. For founders, this signals an opportunity to build or adopt solutions that reduce cloud dependency, particularly in sectors where data sensitivity is a concern. The collaboration with NVIDIA also highlights the role of high-performance local hardware in making on-device AI viable, which could lower barriers for startups needing cost-effective, scalable AI workflows. This trend may push cloud-centric AI providers to offer more flexible hybrid models, giving users greater choice in how and where their data is processed.
Key takeaways
- Founders handling sensitive data can reduce cloud dependency and potential exposure by using local AI processing.
- This launch signals a shift toward hybrid AI models, where critical tasks stay on-device while cloud access remains optional for broader capabilities.
- Startups in regulated industries (finance, healthcare, legal) may find local AI tools more compliant with data privacy requirements.
- The use of NVIDIA’s hardware and open models suggests a growing ecosystem for high-performance local AI, which could lower costs for compute-intensive tasks.
Sources
This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.
About the author
Editor, Lazyfounder
Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.
More stories by Tarun MottliaGet the LazyFounder Brief
Startup, funding and AI news in a five-minute read. Join the early-access list.


