Perplexity Launches Portable Computer on Windows Powered by NVIDIA RTX GPUs

On September 14, 2026, Perplexity officially launched Portable Computer for Windows in collaboration with NVIDIA. Powered by NVIDIA RTX GPUs and TensorRT-LLM acceleration, Portable Computer transforms consumer and workstation PCs into private, autonomous AI desktop workstations capable of executing complex research, document analysis, and multi-step desktop workflows locally with zero cloud API costs.

Local-First Agentic Architecture on NVIDIA RTX Silicon

While traditional AI search engines require continuous cloud API calls that expose private documents and deplete monthly subscriptions, Perplexity Portable Computer executes model inference directly on the user's physical machine. Using NVIDIA TensorRT-LLM and vLLM runtimes optimized for GeForce RTX 40-Series, RTX 50-Series, and RTX Spark systems, the application processes private PDFs, confidential spreadsheets, and source code repositories entirely within local VRAM, achieving up to 1.9x faster inference speeds without streaming enterprise data across the internet.

Core Capabilities & Performance Highlights

• Zero Cloud Credit Consumption: Runs local language models locally on RTX Tensor cores, eliminating per-query subscription limits and cloud latency.
• Hybrid Web & Local Search: Seamlessly merges real-time web citations with offline search across local folder directories and active workspace files.
• Desktop Autonomous Execution: Perceives on-screen context to automate document synthesis, cross-spreadsheet financial analysis, and code debugging.
• Hardware Optimization: Leverages TensorRT-LLM to run quantized 8B, 14B, and 30B open-weights models across GeForce RTX desktop cards, mobile workstations, and upcoming RTX Spark laptops.
• Privacy-Preserving Governance: Air-gapped operational modes ensure sensitive intellectual property and internal emails never leave the physical device.

Accelerating the Shift to On-Device Agentic Work

The release marks a significant transition in personal computing, demonstrating how high-density consumer GPUs can run full-fledged autonomous software agents without relying on hyperscale datacenter servers, delivering private, always-on AI for enterprise researchers and software developers.

Software Specification Sheet