Perplexity Launches Portable Computer Local AI Agent with Nvidia

Perplexity AI Inc. launched Portable Computer, a local-first version of its agentic AI platform developed in partnership with Nvidia. The system allows AI agents to run entirely on a user's local hardware, ensuring that models, files, and workloads remain on the device to eliminate token costs and enhance data privacy. The platform integrates a local model, inference engine, and isolated sandbox, enabling the AI to execute code, read PDFs, and run shell commands locally.

Initially available for Pro, Max, and Enterprise subscribers on Linux and Nvidia DGX Spark desktops, the software requires an Nvidia RTX GPU with at least 24GB of VRAM. Support for Windows is scheduled for September. Users can choose between the open-source Qwen 3.8 27B model or Perplexity's post-trained PPLX 27B, with Nvidia's Nemotron 3.5 Lightning arriving soon. To manage long prompts, the system uses a context compaction feature to summarize text below 100,000 tokens.

While designed for local execution, the system can escalate complex tasks to cloud-based frontier models after requesting user permission. Perplexity research indicates that co-designing the model and agent harness improved performance on knowledge work benchmarks, where the tool achieved an 85.4% score. The launch coincides with reports that Nvidia is considering an investment in Perplexity that would value the company at over $30 billion.