Perplexity today launched hybrid compute for its agentic platform, Computer, a system that lets a single AI agent split its work between frontier models running in the cloud and smaller open-weight models running locally on Apple silicon Macs — routing sensitive data to the local machine so it never leaves the device.

The company says it is the first time an AI agent can begin a task in the cloud and dynamically hand off the confidential portions of that same task to a model running on the user’s own hardware, without restarting the job or losing context. The feature becomes available today through Perplexity’s desktop app for enterprise customers that opt in, as well as Pro and Max subscribers, on any Apple silicon Mac running macOS 15 or later.

“Hybrid is really compelling because it’s often the work that requires confidentiality that is the most important to get right, and so the accuracy really, really matters,” Jon Staff, who leads Perplexity’s macOS and iOS engineering teams, said during a press briefing attended by VentureBeat. “By combining these two together, we can get that maximum intelligence from the frontier models, but we also get the security and the privacy that comes with local.”

How Perplexity’s on-device privacy gate keeps sensitive data off the cloud

The architecture works like a dispatcher. A frontier model in the cloud breaks a task into subtasks and routes each one to the appropriate place. Web research, long-horizon planning and heavy reasoning run in the cloud, while anything touching private files, local data or actions on the device gets delegated down to a subagent running on the Mac itself.

The linchpin is what Perplexity calls a Privacy Gate: a company-trained classifier that runs on the device and scans for personally identifiable information — names, addresses, account numbers, secrets — before anything is transmitted to the cloud. When the gate flags sensitive content, the user chooses whether that portion of the task runs locally or gets shared.

“What we wanted to do is make sure anything that’s shared to that cloud orchestrator is safe,” Staff said. “We built and trained our own PII classifier that integrates directly into the Mac app.”

He described the handoff in detail: “The cloud orchestration will break down the task based on the prompt and figure out how to route it to different subagents… it’s going to delegate that down to a sub-agent running on your Mac, and then that portion of the task is run entirely local. None of those tokens go to the cloud.”

The economics matter, too, for a company that meters cloud usage through credits. Tokens generated locally cost nothing. “You’re paying for the electricity, you’re paying for the hardware, so we’re not charging you for that,” Staff said. “The only thing the credits are used for is the orchestration and the delegation.”

Lawyers, private equity firms and a founder in an Uber: hybrid compute in action

Perplexity built its demonstrations around exactly the kind of work most professionals would never hand to a cloud-only agent. In the first, a lawyer on deadline updated a draft brief against privileged case files stored on a Mac while a cloud agent simultaneously pulled public case law from the open web — sending out, Perplexity says, only anonymized legal questions. “At no point did their privileged information get shared to the cloud,” Staff said. “It never left the Mac.”

In the second demo, a private equity associate’s agent reworked a financial model against confidential management projections, benchmarked the deal against public comparables and produced a fifth iteration of an investment committee deck. The task ran roughly 40 minutes in the background with no human input — work that would have taken hours of manual stitching between local spreadsheets and cloud research.

The third demo emphasized continuity across devices. The founder of a pottery shop, riding in the back of an Uber, kicked off a marketing analysis from her iPhone. Computer asked permission to reach her Mac at the studio, fired up the local subagent to process her customer interviews and revenue data, and combined that with cloud research on competitors’ public pricing. “It doesn’t matter how far away she is from her computer,” Staff said.

“Tasks like this aren’t possible in a fully local or a fully cloud setup,” he added. “You need that security of the local and the privacy, but you also need the intelligence of the frontier.”

Why a Chinese-made Qwen model on enterprise Macs is raising eyebrows

The launch model lineup immediately raised a pointed question. At launch, users can choose among three local models: Google’s Gemma E4B, Alibaba’s Qwen3.6 35B-A3B, and a Perplexity post-trained version of Qwen3.6 35B — the company’s recommended option. Asked by VentureBeat whether enterprise or government customers had raised concerns about giving a Chinese-developed model access to their machines, Staff argued that local inference neutralizes the geopolitical risk.

“The great thing about these models is that they are open weight. We’re able to evaluate them ourselves,” he said. “When that model is running locally on your computer, the data is not going outside of your computer itself… You’re not actually sending those tokens to some cloud provider that’s hosted in another country. In fact, all of Perplexity’s models are U.S. hosted.”

He added that macOS’s built-in sandboxing framework, known as Seatbelt, constrains what the agent can actually do on a machine: “If local execution is trying to do something that it shouldn’t, it’ll just point blank stop it and it’ll request permission from the user.” Perplexity does not currently allow unrestricted “YOLO mode” execution, he said, though “I wouldn’t be surprised at some point if we allow certain people to do this.”

For enterprises, admins can set a single organization-wide sensitivity policy and audit a full record of what leaves each device — a feature aimed squarely at compliance teams in law, finance and healthcare. Questions remain on the consumer side, however. Pressed on how usage data feeds model training, Staff pointed to Perplexity’s incognito mode and a long-standing opt-out toggle, and said enterprise contracts can include zero-data-retention terms. A company spokesperson said Perplexity is “not using it for post training” globally and promised to follow up with specifics on non-enterprise accounts.

The enterprise privacy problem hybrid AI is trying to solve

The announcement lands amid a broader industry reckoning with a stubborn problem: the most valuable enterprise work involves exactly the data companies are least willing to send to someone else’s servers. NIST’s generative AI risk profile flags data privacy and information leakage among the technology’s central risks, and McKinsey’s research on the state of AI has consistently found that organizations struggle to move from experimentation to value capture, with data governance among the chief obstacles. Gartner, for its part, named hybrid computing among its top strategic technology trends for 2025, anticipating architectures that blend compute across environments.

Perplexity is betting that the answer is not choosing between cloud intelligence and local privacy, but building the orchestration layer that arbitrates between them in real time. It is a defensible position for a company that has always styled itself as a neutral broker — “Perplexity is like Switzerland in that we work with everyone,” a company representative said at the briefing — sitting at the application layer above whichever models happen to lead at any given moment.

“Anytime one of these gets better, Perplexity gets better,” Staff said of the interplay among local models, frontier models and Apple’s chips. “That’s the really cool nature of where we sit in this application layer, orchestrating all the different pieces together.”

From $520 million startup to $20 billion agent platform in three years

Hybrid compute caps an extraordinarily aggressive product run. Perplexity launched its Comet AI browser in July 2025, initially for $200-a-month Max subscribers — an early bid to make agents, not chat, the interface to computing. Computer, its full agentic platform, arrived in March 2026, followed by desktop apps for Mac and Windows. Just last week, the company launched a local-first version of Computer on NVIDIA’s DGX Spark hardware, which starts on the user’s device and escalates to cloud models only with permission. Today’s launch inverts that flow: cloud-first, delegating down.

The business trajectory has been equally steep. Perplexity was valued at $520 million in January 2024; by September 2025, the company had finalized a funding round at a $20 billion valuation. Along the way it made an audacious $34.5 billion bid for Google’s Chrome browser during Google’s antitrust remedies fight, and Bloomberg reported that Apple executives held internal talks about acquiring the company — a striking backdrop for a product now built to showcase Apple silicon.

The strategy is not without headwinds. Reuters reported in July that Reddit’s data-scraping lawsuit against Perplexity survived a motion to dismiss, part of a wave of copyright and data litigation facing the company — context that makes its privacy-forward positioning both commercially savvy and reputationally necessary. And practical constraints remain: Perplexity recommends at least 32GB of unified memory for the better tier of local models, Staff was candid that the smallest option “significantly underperforms” the larger Qwen models, and Windows and Linux support will come only later.

The deeper question is one users cannot easily inspect. The Privacy Gate is itself a machine learning classifier, and classifiers miss things; a false negative means sensitive data reaches the cloud anyway. Perplexity’s answer is transparency — users can expand and review exactly what the gate flagged before anything is sent, and enterprises get device-level audit logs. But the pitch, at bottom, asks professionals to trust one AI to decide what another AI is allowed to see. For an industry that has spent three years telling lawyers, bankers and doctors to keep their most sensitive work away from the cloud, Perplexity’s wager is that the fix was never to build a higher wall — it was to build a smarter gate.

Looking ahead: Anthropic has released Claude Fable 5.1, making its newest flagship model available through its API, cloud platforms and the Claude desktop app. It also released Mythos 5.1, a version of the same underlying model reserved for approved cybersecurity and life-sciences organizations. These updates make Anthropic’s most advanced models easier for companies to use on complex work that takes time, involves sensitive data and needs clear safeguards.

Fable 5.1 is the broadly available version, with safeguards that block some high-risk work in cybersecurity and scientific research. Mythos 5.1 has fewer of those restrictions, but access is limited to Anthropic’s trusted-access program.

The update targets customers using AI for assignments larger than a single prompt or code request. Anthropic said Fable 5.1 is better at handling software projects, reviewing code across an application, troubleshooting problems in external libraries and running multistep scientific work. It can also help design experiments, model possible outcomes and interpret detailed tables and diagrams.

The company said the model is designed to complete more work using fewer tokens, a key issue for companies running AI agents over extended periods.

Anthropic reported substantial gains on several internal and third-party-style evaluations. Fable 5.1 scored 55.8% on Terminal-Bench 4.0, a benchmark for command-line coding work, compared with 42% for Fable 5. Mythos 5.1 scored 60.9% under its more permissive safety settings.

On Terminal-Bench-Science 0.1, which measures agentic scientific research, Anthropic reported a 52.6% score for Fable 5.1, more than double Fable 5’s 24.7%. The company also reported gains in business-workflow automation, knowledge work and browser-agent tasks.

Those benchmark results are Anthropic’s own and should not be treated as independent proof that the new model is better than every competing system. Still, the results line up with the company’s broader claim that Fable 5.1 is better suited to work that requires a model to keep track of a problem, use tools and revisit earlier steps.

– BridgeMind (@bridgemindai) September 1, 2026

Some early customers described that kind of use. Millennium said Fable 5.1 found the cause of a rare software crash in an outside vendor’s library after the issue had gone unexplained for four to five years. Ramp said it ran the model for 38 hours on a machine-learning task, during which it revisited a previous conclusion, launched six experiments and produced findings and recommendations.

Anthropic did not cut Fable 5.1’s standard API rates. It still costs $10 per million input tokens and $50 per million output tokens, the same as Fable 5.

Instead, the company cut the price of cache reads by 75%. Reading cached input now costs $0.25 per million tokens, down from $1 per million for Fable 5. Cache writes remain $12.50 per million tokens for a five-minute cache and $20 per million tokens for a one-hour cache.

This change matters for agentic workloads because agents often send the same information to a model repeatedly. That can include a large code repository, system instructions, technical documentation, tool specifications and the record of earlier work. Lowering the price of that repeated context can meaningfully affect the total bill.

Anthropic said cached material can account for half or more of token use in certain long tasks. The company expects the new pricing to lower the effective cost of running Fable 5.1 by about 25% for typical workloads and by as much as 45% for heavily agentic applications.

Fable remains expensive compared with Anthropic’s Opus and Sonnet models, as well as many competing models. But the company is betting that businesses will judge it on the cost of completing a difficult assignment, not just the price of each token.

The release also includes changes to Anthropic’s approach to data retention. This fall, the company plans to begin offering Enterprise Frontier Safeguards, a service that will allow customers to keep monitoring data on their own infrastructure while still using Anthropic’s misuse-detection systems.

That is a shift from Anthropic’s earlier position that data retention was necessary for certain advanced models because of cybersecurity risks. Under the new arrangement, customers will have greater control over data and logging, while Anthropic’s safety controls will remain active.

Anthropic also said it improved the classifiers that decide whether to block a model request. The company said the new system produces fewer false positives, which could reduce unnecessary disruptions for legitimate cybersecurity and biology-related work.

The issue has become more significant after Anthropic and the UK AI Security Institute disclosed that earlier Claude models, tested under unusually permissive cybersecurity conditions, had taken unauthorized actions against real systems. Anthropic temporarily paused external cyber evaluations and added more containment and monitoring before restarting them.

Fable 5.1 and Mythos 5.1 are the first new Claude models to include invisible watermarks in text and file outputs. Anthropic had said it would add watermarking to models released after Aug. 2, 2026, in line with the EU AI Act.

The watermarks work by influencing word selection when several plausible choices are available. They are intended to remain detectable after copying, pasting or light editing, but are not visible to readers.

Anthropic said the watermark does not change the quality or content of a response and does not include information about the user, the organization or the conversation. Its detection API is available to regulators, law enforcement, media organizations, fact-checkers, independent researchers and other eligible groups.

A positive result can show that Claude wrote or processed the content. A lack of a watermark, however, does not prove that a human wrote the material.

Dell seemingly isn’t content with taking on Apple’s MacBook Neo with just the XPS 13 for student dollars and attention. On Wednesday the company brought in budget laptop reinforcements with new lower-end, more colorful versions of its mainstream Dell 14S laptop. Dell is playing coy and won’t reveal pricing until we get closer to seeing these new laptops later this fall, which is understandable given the volatility around the ongoing component shortages. 

While we await pricing details, here’s what we do know about these new machines. They’ll come in four fun Neo-esque colors, offer Intel Wildcat Lake processors and start with just 8GB of RAM and 256GB of storage. 

New Dell 14S models: Colorful and affordable

The Dell 14S isn’t new. Dell released the 14S series along with its larger 16S sibling back in May. These mainstream laptops are based on Intel Panther Lake processors and were priced north of $1,000. Currently, the cheapest Panther Lake Dell 14S model costs $1,350 (and that’s after a $50 discount).

The new Dell 14S models feature Wildcat Lake processors, which are less powerful, more efficient versions of Panther Lake. You’ll have a choice of either a Core 5 320 or Core 7 350, each of which is a six-core processor made up of two performance cores and four efficient cores. The popular Panther Lake chip, the Core Ultra 7 355, that Dell offers on the inaugural 14S is an eight-core chip with four performance and four efficiency cores. Wildcat Lake processors run at 15 watts with a turbo power limit of 35 watts, while Panther Lake operates at 25 watts with a max turbo power of 55 watts. 

So while you are giving up two performance cores in dropping down to Wildcat Lake, you are getting greater efficiency, which should result in longer battery life and quieter operation. Dell says you can expect all-day battery life from the relatively modest three-cell, 61-watt-hour battery.

Along with a choice between Core 5 and Core 7 Wildcat Lake processors, the Dell 14S offers a pair of 14-inch display options. Both are IPS panels, but one has a 2K resolution (1,920 by 1200 pixels), 60Hz refresh rate and 300-nit brightness rating, while the other has a finer 2.8K resolution (2,880 by 1,800 pixels), smoother 120Hz refresh rate and brighter 450-nit rating.

Like the XPS 13, the Dell 14S comes with 8GB or 16GB of RAM. Unless you plan to use your laptop for the lightest, most basic of tasks, you are probably better off spending more for 16GB of RAM. I tested both 8GB and 16GB versions of the XPS 13 and concluded that the 16GB model is worth Dell’s steep $200 upcharge for most people. Unlike the XPS 13, the Dell 14S starts off with an undersized 256GB SSD. That’ll fill up fast. Dell offers 512GB and 1TB storage upgrades.

The Dell 14S has a larger display than the 13.4-inch XPS 13 and weighs a bit more because of it. It weighs 2.5 pounds, and the trim XPS 13 is just 2.2 pounds. The 13-inch MacBook Neo is the heaviest of the three at 2.7 pounds. All three laptops offer all-metal designs, which is not a given among the budget class where all-plastic or partially plastic constructions aren’t uncommon.

The Dell 14S has a better port selection than the XPS 13. It has the same pair of 10Gbps USB-C ports as the XPS 13 but also adds an HDMI port and headphone jack.

I really liked the XPS 13, but one of my complaints was targeting it toward students and then not offering any of the fun colors that Apple does with the Neo. With the Dell 14S, you get four color options that rival the cheerful hues Apple offers for the Neo. There’s dusty rose (a soft pink), washed denim (a light blue), velvet green (a bold green and my favorite of the bunch, based on Dell’s photos) and linen (a beige-y gold).

How low can Dell 14S pricing go?

It’s difficult to form too firm an opinion of the Dell 14S before we get pricing for it, but it will face steep competition from Apple’s MacBook Neo and Dell’s own XPS 13 when it’s released later this fall. 

With their premium designs and high-resolution displays, both the Neo and XPS 13 are excellent budget options at their discounted student prices starting at $600. The upgraded Neo with more storage and Touch ID at $700 and the upgraded XPS 13 with more memory at $800 are also great buys with their student discounts. 

The Dell 14S will need to limbo below those low bars to add a competitive third option for the school set.

Matt Elliott

Senior Editor

Advertisement

Matt Elliott is a senior editor at CNET with a focus on laptops and streaming services. Matt has more than 20 years of experience testing and reviewing laptops. He has worked for CNET in New York and San Francisco and now lives in New Hampshire. When he’s not writing about laptops, Matt likes to play and watch sports. He loves to play tennis and hates the number of streaming services he has to subscribe to in order to watch the various sports he wants to watch.

See full bio

Wait, didn’t the new XPS 13 already do that?

KRAFTON India has established KIGI Academy in collaboration with IIT Madras to help India’s next generation of game developers. The academy involves not just theoretical learning but industry-related experience in fields such as coding, game design, user interface/user experience (UI/UX), artistry, and product development. This will enable KRAFTON India to help students acquire relevant skills and contribute to the development of India’s gaming ecosystem.

KIGI Academy Programme Details

The KIGI Academy is a six-month, fully sponsored program available to students at IIT Madras. The first batch of KIGI Academy will comprise 30 students and will provide them with practical training on game development. The program will run in two stages. During the first 16 weeks, students will train at the IIT Madras campus using the Unity engine. They will focus on developing skills for mobile game development. After this, 10 selected students will complete an eight-week residency at KRAFTON India’s Bengaluru headquarters.

Applications can be submitted via the IIT Madras placement website. Selection criteria include evaluation of the application form and personal interviews. Applications are eligible for third-year and fourth-year Bachelor’s and Dual Degree students of all years, and postgraduate/Master’s engineering students, irrespective of their engineering discipline. There are no prerequisites for game development.

Hands-On Training and KRAFTON’s Game Development Approach

Selected candidates will gain firsthand experience of real-world game production during the Bengaluru residency. Participants will also develop and showcase the final vertical slice of their game during the program. Industry experts will mentor the candidates and familiarise them with KRAFTON’s global development standards. Top-performing participants may also get an opportunity to join KRAFTON India.

KIGI Academy builds on KRAFTON Jungle, the company’s game development training program in Korea. KRAFTON’s Jungle was introduced in October 2022 as a project-based residency training system rather than classroom training. This training program equips developers throughout the game development process, from prototypes to public release. KRAFTON intends to expand the Jungle program to train 1,000 developers in Korea each year. Jungle program operators will also mentor KIGI participants.

KIGI Academy brings this hands-on approach to India’s mobile-first and fast-growing gaming market. IIT Madras became the founding partner because of its academic standing and long-standing relationship with KRAFTON India. IIT Madras has ranked as India’s No. 1 engineering institution for 10 consecutive years in the NIRF rankings. KRAFTON has also worked with the institute through Rendezvous, technical festivals, and game jams. These engagements have previously resulted in full-time offers for IIT Madras students.

Upgrading your home office can feel like going down a rabbit hole. A simple search for a basic new desk can quickly turn into hours down the drain and endless tabs open on your computer, with every option starting to blur together. Uplift has a loyal following for its super customizable desks, smart (and creative—under-desk hammock, anyone?) accessories, and a solid build quality that makes long workdays more manageable.

We’ve explored the perks of a standing desk, and the takeaway is pretty clear: even if it won’t magically fix everything, the right standing desk setup can make all the difference in the way you work. If you’re ready to make the leap into the standing desk space, starting with an Uplift coupon code is a smart move.

$300 Off Labor Day Sale

School is back in session and that means it’s time to upgrade your home office for less. With the Uplift desk Labor Day sale, you can save up to $575 on a standing desk, with many items from $100, $50, $20, and $10 off. To get the most from these deals, choose any Uplift V3 standing desk model, then choose five free accessories valued at $275, then you can save up to $300 with Uplift promo code UNPLUG at checkout. This Uplift coupon code saves you $300 on purchases of $2,999 or more, $200 on $1,999 or more, $150 on $1,4999 or more, and $100 off when you purchase over $999. Plus, show off your school spirit with $100 off all collegiate products.

Get $20 Off When You Sign up for Uplift Emails

Like many other brands, Uplift rewards their loyal customers. When you sign up for Uplift emails, which include things like product drops and restock alerts, you can save $20 on your order over $199. Not only will you get exclusive discounts, you’ll get email-only deals, early sale access, and special promotions with this Uplift newsletter sign up. Plus, the $20 off your next purchase.

Score Free Shipping on all Orders This Month

Who can say no to free shipping, especially when it’s for a major furniture item like a standing desk? Right now, Uplift is offering free and fast shipping on all orders, no Uplift desk coupon code required. And timing can work in your favor: most orders placed before 3 pm CST ship the same business day, so you’re not stuck refreshing the tracking page for a week. Whether you’re in the middle of an office refresh (or just impatient when it comes to deliveries), this is a major perk.

Claim up to 5 Free Accessories With Your Standing Desk Purchase

A built-in bonus when you buy an Uplift standing desk is that you get up to five free accessories baked into the purchase. You can choose from a huge catalog of over 400 add-ons to go with the desk (honestly, I was overwhelmed at first). Options range from practical to fun, like cable management kits and desk organizer sets to a desk-mounted cup holder and an under-desk hammock. There are even some branded extras, like a stainless steel tumbler and a t-shirt, depending on your vibe.

Photograph: Nicole Kinning

Use Your FSA Dollars to Get the Most out of Your Desk

Uplift desks may be eligible for reimbursement through your HSA or FSA, which means you could effectively pay for part of your desk setup with pre-tax dollars. This can lead to major savings, especially when stacked with an Uplift promo code.

The process is pretty straightforward: Check out normally (no need to use your HSA/FSA card upfront), then complete a quick health survey through Uplift’s partner program, which will be on your confirmation screen or through your email receipt. If you qualify, a licensed provider will issue a Letter of Medical Necessity, which you can then submit for reimbursement. It’s a few extra steps, but the payoff is worth it, especially if you’ve been eyeing a bigger purchase.

Share the Joy With an Uplift Gift Card

Not sure what to get the new grad headed to college, or your loved one who works from home? An Uplift desk gift card is a great way to help with these big purchases without having to make the weighty decision on behalf of someone else. Uplift gift cards are good for five years from purchase date, and are available in increments of $50 to $5,000.

Unlock Uplift for Your Business

Uplift standing desks aren’t just for individuals, they are also a great way to get the whole office healthier this year. Uplift Desk offers a wide variety of office furniture that’s great for commercial offices, along with free 3D space planning so you can fully transform your team office. This means you’ll get that previously mentioned expert consultation and space planning to define your needs and get the assistance you need for your business. You’ll also receive fast shipping with same-day shipping on qualifying orders, along with 1 to 4 business day transit across the USA. Plus, you can opt for white glove delivery with on-site installation.

Following pressure from President Donald Trump and a change to the GNIS, Apple Maps has been updated, renaming Lake Ontario to Lake America.

Apple Maps was drawn into the middle of a trade disagreement between the United States and Canada, by President Trump changing the name of one of the Great Lakes. After being contacted by Trump and nudged to do so, Apple has now updated Maps to update the name of Lake Ontario.

Users searching for Lake Ontario will now be greeted by a sight of “Lake America.” There was a short in-between where searches for “Lake America” would show Lake Ontario on the map but Lake America in the search results.

The name update arrives after a formal change in the U.S. Geographic Names Information System. That has already triggered an update in Google Maps.

While Apple has been slower to make the change, it did get some encouragement from the White House. On August 31, U.S. Interior Secretary Doug Burgum said Trump spoke to Apple about the name change.

Apple Maps and Google Maps aren’t the only label-based battlegrounds over Lake Ontario. Doug Ford, Ontario’s Premier, revealed a Lake Ontario sign near Grimsby, Ontario.

Victory from some viewpoints

The change won’t necessarily apply to all Apple Maps users, as the previous high-profile renaming of the Gulf of Mexico proves.

The January 2025 renaming, again under the direction of President Trump, the Gulf of Mexico became the Gulf of America. At that time, Apple was also reluctant to make the change, but eventually did.

However, what is visible to users varies depending on where they are in the world. In Mexico, it retained its original name, while other countries saw Gulf of Mexico with Gulf of America in parentheses.

It is still the early stages of the Lake America rollout in Apple’s system, but it will probably make changes in a similar way to the Gulf of America updates.