Portable Computer is now available on Windows PCs with @NVIDIA RTX GPUs. Run the harness, agents, and models locally on your PC. Work with local files and connected apps without sending tasks to the cloud. Use frontier cloud models when needed.

We’ve also added support for local MCP and scheduled tasks. Connect Portable Computer to your own tools and app integrations through local MCP. Give Computer recurring work and schedule it to run on your PC while you’re away.

Get started with Portable Computer in the Perplexity Windows app. On-device inference requires an NVIDIA RTX GPU with 24GB of VRAM or higher.

@nvidia nvidia's out here selling the shovels and perplexity's selling the map to the shovel store

@nvidia the underrated part is converting fat mcp schemas into compact clis. local models were choking on tool defs before they could do the work.

@nvidia 🤮🤮🤮

@nvidia pro plan ? api ?

@nvidia Not free

@nvidia Local-first with cloud as fallback is how it should work. Honestly, the real test is whether the local mode stays fully functional or quietly gets nerfed to push people to a cloud tier. Owning your own compute = sanity.

@nvidia This is the direction I want local AI to go. But 24GB VRAM still puts it well outside the average Windows PC. The really interesting part is local by default, with the option to use a cloud model only when you actually need it.

@nvidia The connected-app boundary is the bit I'd test first, especially with local files

@nvidia That's incredible

@nvidia Mac support soon?

@nvidia Waiting for you to add M5 Ultra support up to 512GB please

@nvidia Does this work on DGX Spark

@nvidia 24GB VRAM 门槛把大多数游戏卡挡出去了。真正能用上的是本机 harness + 本地 MCP + 定时任务这一套，文件不用上云还能排过夜活。

@nvidia Local harness + agents on a Windows RTX box is the right packaging. The part operators care about is how you freeze a working loop (tools, memory, stop conditions) so it survives a reboot — not another chat window with a GPU.

@nvidia Two modes: harness on the GPU in the box, cloud model only when you send the task.