Perplexity has officially announced an on-device version of its AI agent framework, dubbed Portable Computer. The deployment allows artificial intelligence models to execute entirely on a user’s own machine, removing dependencies on cloud infrastructure while significantly enhancing data privacy.
Previously, hybrid setups handled routine tasks locally on entry-level hardware like the Apple Mac Mini while offloading compute-heavy jobs to the cloud. Portable Computer replaces that cloud layer, shifting intensive workloads directly to local components. However, this transition requires enterprise-grade hardware. At launch, compatibility is limited to the Nvidia DGX Spark and Linux workstations equipped with high-end Nvidia RTX graphics cards.
The hardware entry point represents a substantial shift in capital expenditure. While a standard Mac Mini begins around $900, an Nvidia DGX Spark system commands a price tag of roughly $5,199—nearly six times higher.
Key Benefits and Hardware Specifications
- Zero Monthly Token Fees: Heavy enterprise and power users can replace recurring cloud API charges with a one-time upfront hardware investment.
- Hybrid Cloud Option: The framework processes all standard workloads locally, querying external cloud servers only when explicitly authorized for supplementary compute.
- Supported AI Models: Users can deploy either Alibaba’s Qwen 3.8 27B or Perplexity’s fine-tuned PPLX 27B model, with support for Nvidia Nemotron 3.5 Lightning scheduled for a later update.
- App Integrations: Despite running locally offline, the assistant automates tasks across everyday platforms, successfully generating data reports for Slack and interfacing directly with Google Drive, Gmail, and GitHub.
This release underscores a deepening relationship with hardware partners. Nvidia is reportedly exploring a strategic investment in Perplexity that could elevate the AI startup’s market valuation to $30 billion, accelerating the development of dedicated local inference solutions across the US technology sector.
Filed in . Read more about AI (Artificial Intelligence).