Perplexity AI Launches 'Portable Computer' for Local AI Workflows
Perplexity AI has launched "Portable Computer," a local-first version of its multi-model agent that runs on-device using Qwen 3.8 27B and its custom PPLX 27B models. The application initially supports Linux-based NVIDIA DGX Spark systems, with Windows and GeForce RTX support coming soon. This release marks a shift towards hybrid local-cloud AI workflows, allowing users to process sensitive tasks privately on-device while falling back to the cloud only when necessary. It addresses growing data privacy and latency concerns in enterprise and personal AI assistant usage. The on-device PPLX 27B model is a post-trained version of Alibaba's Qwen 3.8 27B, achieving an 85.4% score on real knowledge work benchmarks. Additionally, Perplexity is optimizing a fine-tuned version of NVIDIA's Nemotron 3.5 Lightning model to achieve ultra-fast response times.
## BACKGROUND
Traditional AI agents rely heavily on cloud-based Large Language Models (LLMs), which can raise privacy concerns for sensitive data and introduce network latency. Local or edge AI running on specialized hardware like NVIDIA's DGX Spark or RTX GPUs allows models to run directly on a user's machine, keeping data local and secure.