Perplexity Portable Computer Pricing and the DGX Spark Bill

Portable Computer doesn't have a standalone price. At Layer3Labs, we build and operate AI systems inside other companies, and the price on the box is rarely what determines whether a rollout makes sense.

Perplexity includes Portable Computer with four paid plans: Pro, Max, Enterprise Pro, and Enterprise Max. Subscribe to any one of them, and Portable Computer is included.

The bigger expense is the hardware. Because Portable Computer runs the model locally, you'll need either an NVIDIA DGX Spark or a Linux PC with an NVIDIA RTX graphics card and at least 24GB of video memory. As of 27 August 2026, NVIDIA listed the DGX Spark for $4,699 on its marketplace. Before budgeting, check NVIDIA's DGX Spark product page for the latest price.

There's also a third, optional cost. For difficult tasks, Portable Computer can ask a frontier model in the cloud for help. Perplexity puts that step at roughly $0.415 per task.

So the real cost has three parts: the subscription, the machine, and any cloud escalations. Because Perplexity's plan prices change often, we link to its pricing page rather than publish a figure that may quickly become outdated.

What Does Perplexity Portable Computer Cost?

Perplexity charges nothing extra for Portable Computer itself.

It comes with four paid tiers: Pro, Max, Enterprise Pro, and Enterprise Max. A subscriber on any of those can install it, which means an existing Pro subscriber pays zero additional dollars for the software side of this.

That splits the price question in two. What you pay Perplexity every month, and what you pay NVIDIA once.

We do not print Perplexity subscription figures here on purpose. They move, and a stale number in a budget spreadsheet costs more trust than it saves time. Perplexity's pricing page carries the current amount for every tier.

• Portable Computer software: no separate purchase, it comes with a qualifying tier
• Subscription: Pro, Max, Enterprise Pro, or Enterprise Max, at whatever Perplexity currently charges
• Hardware: a DGX Spark at $4,699, or a Linux PC with a 24GB or larger NVIDIA RTX card
• Cloud advisor: approximately $0.415 per escalated task, a figure Perplexity reported on its own benchmark

Which Perplexity Tier You Need for Portable Computer

All four paid tiers carry Portable Computer: Pro, Max, Enterprise Pro, and Enterprise Max.

Perplexity has not described a different Portable Computer for any one of them. The same local agent runs on the same hardware whichever of the four you hold, so the local agent itself is not a reason to pick one tier over another.

That makes the Perplexity Pro vs Max question a cloud-side question. Decide it on what each tier adds beyond Portable Computer, and read the current price of each on Perplexity's pricing page rather than on a figure quoted anywhere else.

One caution for teams. Enterprise seats are bought per person, so a five-person team running Portable Computer pays for five seats, while the DGX Spark it runs on is bought once. Put the seat price you are quoted and the machine price on the same line before you assume which half costs more.

• Pro: includes Portable Computer
• Max: the same Portable Computer access, so pick it for what else it adds
• Enterprise Pro and Enterprise Max: the business tiers, also included
• No tier sells Portable Computer as a standalone add-on

What the NVIDIA DGX Spark Costs

NVIDIA listed the DGX Spark at $4,699 on its marketplace as of 27 August 2026.

The Spark is a Grace Blackwell GB10 desktop machine built for local model work. It arrives ready to run Portable Computer, so there is no parts list to assemble and no driver hunt before the first task.

Sparks can be linked over shared memory in groups of two, four, or eight. A team that outgrows one box adds another instead of replacing the first, which turns a rebuild into a purchase order.

Prices on hardware like this move. NVIDIA listed $4,699 on one day in August 2026, and the DGX Spark product page carries both the current figure and the order link. We have not run a DGX Spark ourselves, so everything here about the machine comes from what NVIDIA and Perplexity publish.

• $4,699 listed by NVIDIA on 27 August 2026, verify current pricing before ordering
• Grace Blackwell GB10, sold as a finished desktop machine
• Links in groups of 2, 4, or 8 over shared memory for larger models
• Ordered through NVIDIA's own marketplace listing

The RTX PC Path and Its 24GB Floor

A Linux PC with an NVIDIA RTX graphics card is the second route. It adds nothing to the bill if you already own a card that qualifies.

The reported floor is 24GB of video memory, which means an RTX 3090 or newer. 32GB is cited as the official recommendation, so a 24GB card sits at the edge of what runs rather than in comfortable territory.

Linux is the only platform at launch, and Perplexity has stated Windows support for September 2026. There is no Apple silicon support and none on the roadmap, so a Mac stays out at any budget. If you run a Mac and want a local model anyway, a different runtime is the route: our Ollama guide covers what a Mac can do instead.

Buying a card new can cost more than this route appears to save. Street prices for 24GB and 32GB cards swing week to week. This path beats a Spark only when the card you want is cheap on the day you buy. A fresh build adds a processor, memory, storage, a power supply, and a case on top of the card. Our local AI hardware calculator sizes the memory you need before you shop.

• 24GB of video memory is the reported minimum, meaning an RTX 3090 or newer
• 32GB is the cited recommendation, so 24GB is the floor rather than the target
• Linux only at launch, with Windows stated for September 2026
• No Apple silicon support, and none planned

When the $0.415 per Task Figure Applies

The $0.415 per task figure applies only when Portable Computer escalates a hard task to a frontier model in the cloud, with your permission.

Perplexity's own Terminal Bench 2.1 coding results put Portable Computer at 59.6% running fully locally and 73.0% with the cloud advisor, at approximately $0.415 per task. Those are Perplexity's numbers on Perplexity's benchmark, not independent testing, so treat the 13.4-point gain as a vendor claim until someone else reproduces it.

The advisor returns text guidance only. It cannot touch local files or local tools, which means the money buys reasoning rather than actions taken on your machine.

Work the local models finish alone carries no per-task figure at all. Web searches and connector calls also leave the device. Perplexity has not published a separate charge for those, so the advisor is the only priced escalation we can point to.

Across the routine automation we run on our own portfolio of sites, teams routinely forget to model the per-task line. A rate that reads as small next to a $4,699 machine stops reading that way at volume. 300 escalated tasks a week is about $124.50, or roughly $6,474 over a year. That is more than the Spark. Perplexity has not published how, or whether, this cost reaches a subscriber's bill, so confirm what your tier covers on Perplexity's pricing page before you plan around heavy escalation.

• Local only: 59.6% on Terminal Bench 2.1, no per-task figure published
• With the cloud advisor: 73.0%, at approximately $0.415 per task
• Escalation requires your permission and returns text guidance only
• All of these are Perplexity's own reported numbers

What Perplexity Pricing and Hardware Add Up to in Year One

On the DGX Spark path, year one is $4,699 for the machine, twelve months of a qualifying Perplexity subscription, and whatever escalation you approve.

Year two drops to the subscription and the escalation, since the hardware is paid for. Spread across three years, the Spark works out to about $1,566 a year before anything else is counted.

On the RTX path, if you already own a 24GB or larger card, year one is the subscription and the escalation. Nothing else.

If you have to buy the card, the answer depends on the price you find that week, and a fresh build adds the rest of the parts on top. Here is how the three routes compare on the lines we can price.

Most buyers watch the hardware price, and the repeating costs move the total more. Hardware is a single payment you can spread over years. The subscription and the escalation repeat forever, and a two-person team on Enterprise seats can pass the price of a Spark within a couple of years.

Who Should Not Spend This Money

Anyone whose graphics card falls under 24GB of video memory, and who has no other use for a new one, should skip this.

A Mac owner is out at any budget, because Apple silicon is unsupported and not on the roadmap. A Windows-only shop can simply wait, since Perplexity has stated support for September 2026 and waiting costs time rather than money.

Anyone whose work leans on hard reasoning should read the limits before spending. Perplexity says the compact models trail frontier models meaningfully on difficult tasks, and the advertised 260,000 token context is reported to degrade past roughly 100,000 tokens. Paying $4,699 to run a model that hands the hard half back to the cloud is a poor trade if the hard half is most of your work.

Two things would change our answer. A real cut to the DGX Spark price would make the three-year arithmetic work for much smaller teams. So would a plain rate card from Perplexity tying the $0.415 figure to an actual subscriber charge. Right now nobody can quote the largest number in this budget.

Before you commit, put three lines in a spreadsheet: the DGX Spark price from NVIDIA, a year of Perplexity pricing from the tier list, and your escalated tasks times $0.415.

• Skip it if your card is under 24GB and you do not want a new one
• Skip it on a Mac, since Apple silicon is unsupported
• Wait if you are Windows-only, with support stated for September 2026
• Reconsider if hard reasoning is most of your workload