Today we’re launching Portable Computer on @NVIDIA DGX Spark. Portable Computer is a fully local version of Perplexity Computer, where the entire runtime: orchestrator LLM, subagent LLM, agent harness all run on your local hardware. No cloud dependency.

When a task needs frontier reasoning, the orchestrator asks for approval before routing it to a frontier model in the cloud. It's user-gated, PII-flagged, and text guidance only. Your sensitive documents stay on your device.

Portable Computer runs a post-trained and locally installed PPLX 27B model. Qwen 3.8 27B is also available, with support for NVIDIA Nemotron 3.5 Lightning coming soon.

Portable Computer is available today for all Perplexity Pro and Max subscribers with an @NVIDIA DGX Spark. Get started:

@nvidia This app looks familiar 🫣

@nvidia This is awesome. Going to try this on my DGX Spark this week. @AravSrinivas

@nvidia Massive for Local AI

@nvidia good shit perplexity coming back. too bad hardware for local ai is still so bad, cant really use qwen 27b for anything serious. we are all living in a bubble and 2-3 years early lol

@nvidia 🔥

@nvidia Wow

@nvidia You guys are still alive? 🥹

@nvidia Doesn’t look like access to the local machine is available remotely. I.e. you’d have to be on the Spark’s desktop to use it. Any plans to allow access on a MacBook or iPhone via the web and mobile apps?

@nvidia This is a big deal, huge congrats to the whole team 🔥

@nvidia I don't like this trend: "Fully local AI"... With a subscription.

@nvidia It’s a pity this requires having a display attached to the DGX Spark. It would be awesome if I could use Perplexity Computer on my Mac while also having DGX Spark being available via Tailscale inside Perplexity Computer.

@nvidia very cool work!!

@nvidia @grok dumb it down for me - will it run on Apple Silicon? Or specially just a DGX Spark?

@nvidia This is great. Could you guys also add the option for a local inference server on the Computer Mac App? Many people use the Spark as an inference server and don’t work on to directly (like me).

@nvidia Will this eventually reach macOS as well?

@nvidia I run local inference on a DGX Spark daily. The hardware is ready, the tooling isn’t. Open source this stack and Perplexity owns local AI overnight. Keep it closed and it’s just another demo.

@nvidia Wow. And I actually have a DGX Spark. Pretty exciting!

@nvidia Sick, gonna put this in a VM tho

@nvidia Did not see Perplexity taking this leap. Interesting to see how the distribution for this would work!

@nvidia I dreamed a dream in time gone by

@nvidia The DGX Spark can run larger and more capable models than 27b-class. It would be great if perplexity computer would let us pick which model to run, and have support for larger models across multiple sparks (like Deepseek v4 Flash)

@nvidia NICE

The local-first design with gated cloud escalation is the part that actually matters for real agentic work. Keeping files and context on-device while still being able to pull frontier reasoning when needed feels like the pragmatic middle ground between pure cloud dependency and pure offline limits.

@nvidia open source it so we can run it on our own dgx sparks, or any other hardware!

@NVIDIARTXSpark @nvidia Rad. Can this end up on mac too?

@NVIDIARTXSpark @nvidia @CardilloSamuel looks like cloudless? 👀

@nvidia Nice move, Perplexity. Fully local agent on DGX Spark with optional cloud only when you approve it. Privacy + power done right.

@nvidia Congrats team!

@nvidia This is the direction AI agents need to go. A fully local computer agent means more privacy, lower latency, and less dependence on cloud infrastructure. The personal AI computer is getting real.

@nvidia If it's all local, no cloud, why does it need a paid Perplexity plan?

@nvidia Wow OK things getting Hot in IT i see, so happy to be in this field 😍

@nvidia No thanks, I'm not wasting my Spark RAM on Perplexity. Funny how an influencer mentions Perplexity a few days ago and suddenly this. Totally organic, right...

@nvidia Yo, local agents with privacy intact is huge for real

@nvidia Big fan of Perplexity with Max Plan here. How would this be better than Hermes? @perplexity_ai @AravSrinivas