Perplexity app (Personal Computer and Portable Computer)
Perplexity's desktop app, whose Mac hybrid mode offers local models downloaded in the app beside cloud models.
Our description, from the vendor's documentation, as of 2026-10-20; reasoned from those pages, not observed running.
Also found under the names: Perplexity Personal Computer; Perplexity Computer (Mac app); Perplexity (Computer, Mac app); Perplexity Computer (Mac app, Hybrid Compute).
Runs onThe places the vendor's own pages say the product runs: your machine, a container environment you run, or a cloud service the vendor runs. It is not a limit.
Models runWhere the vendor's own pages say the product's models run, not a limit: your machine, a container environment you run, the model company's cloud (servers of the company that makes or serves the model), or a cloud provider's model service. For a cloud, read the data and compliance lines of the company running it, linked by its name.
Several providers in one sessionOne session can use models from different providers, for example one agent on one maker's model and another on a second maker's. Switching the single model for the whole session does not count on its own. A tool that is itself a gateway or library routing to the providers its page names counts.
Several endpoints built inThe tool itself can hold more than one provider or address, so nothing has to sit in front of it to reach different providers.
Local and cloud togetherA local model (on your machine or in a container environment you run) and a cloud model (in its model company's cloud or a cloud provider's model service) can both be used in the same session or setup.
Use a local model: How to set it up(www.perplexity.ai page, read 2026-10-22) Settings the guide names: Hybrid, Local Inference, On-device models, PPLX 27B.
Use a local and a cloud model together: How to set it up(www.perplexity.ai page, read 2026-10-22) Settings the guide names: Hybrid, Local.
Compared with similar tools
Same role (harness / agent client) at the same layer (Coding agents and assistants). Up to 4 others, in alphabetical order; nothing is ranked.
Product
Several providersOne session can use models from different providers, for example one agent on one maker's model and another on a second maker's. Switching the single model for the whole session does not count on its own. A tool that is itself a gateway or library routing to the providers its page names counts.
Several endpointsThe tool itself can hold more than one provider or address, so nothing has to sit in front of it to reach different providers.
Model on your machineThe tool can use a model running on your own computer, for example through Ollama, LM Studio, oMLX or llama.cpp.
Local and cloudA local model (on your machine or in a container environment you run) and a cloud model (in its model company's cloud or a cloud provider's model service) can both be used in the same session or setup.
Perplexity app (Personal Computer and Portable Computer)
Words in the cells: YesA page we read says so., PartlyA page says part of it, for example separate models per agent without saying the providers can differ., Via a proxyOnly with a proxy, or a local server that imitates a cloud provider's interface, in front of the tool., NoA page we read says the product does not support it., ClaimedA vendor page says so on a marketing or overview page, and we found no guide showing how to set it up., Not foundWe read pages for it and found nothing there. It says nothing about whether the product has the feature., n/aThe question does not apply to the product, for example because it is a model server and not a tool that connects to models., BlankNothing is published for it yet.
Data and compliance
Compliance: Not checked yetWe have not read a page of the right kind for this question yet.: no page that could answer it is recorded as read.
Data: Training use, default retention, zero-retention option and data residency not checked yetWe have not read a page of the right kind for this question yet..
Model and where it runs: Installed: a desktop application; model provider and where it can run not checked yetWe have not read a page of the right kind for this question yet..
Details: What the maker's own pages state, not a compliance determination.
What the maker's own pages said on the dates shown, not a compliance determination.
Our summary of the vendor's pages as of the dates shown on each line (2026-10-19 to 2026-10-22); confirm on the vendor's page.
Cloud models
Perplexity's changelog says Computer tasks can use Perplexity's hosted service together with a model running on the user's own Mac. (www.perplexity.ai page, read 2026-10-20)
What you set up
Perplexity's help page gives the Mac steps for a hybrid task: download and open the app, download a local model, open the model selector and choose Hybrid, then choose a local model and a cloud model for the task. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Personal Computer help page gives the Windows steps for a local model: install the app and sign in, open Settings from the Menu, choose Local Inference, download PPLX 27B under On-device models (Qwen 27B is the alternative for AMD Ryzen AI Max hardware), then pick the model under Local in the composer's model menu; setup needs roughly 20 GB of spare disk space. (www.perplexity.ai page, read 2026-10-22)
Models from different providers in one session
Computer builds teams of agents that span 20+ frontier models and automatically picks a combination of models for each task. (www.perplexity.ai page, read 2026-10-19)
On Mac in hybrid mode, the user chooses both a local model and a cloud model for the task. (www.perplexity.ai page, read 2026-10-19)
Perplexity's Model Council help page says the default lineup draws one frontier model from each of Anthropic, OpenAI and Google. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Model Council help page says that in Computer you pick from two to eight models, or let Computer choose the default lineup. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Model Council help page says that inside Computer, Model Council is a skill available on all paid plans. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page lists the Search models you can choose by provider (OpenAI, Google, Anthropic, Moonshot AI, Z.ai, xAI and NVIDIA) as of October 17, 2026, with some of them Max-only on the consumer plans. (www.perplexity.ai page, read 2026-10-22)
Perplexity's effort-mode help page says the Windows and macOS desktop apps do not offer effort mode and keep the regular model picker. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says Computer has a model selector and availability rules of its own, apart from Search. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Personal Computer help page says Personal Computer is the superset of Computer: it does everything Computer does and also has access to your files, apps and desktop. (www.perplexity.ai page, read 2026-10-22)
Several endpoints, built in (no proxy)
Not found in the pages we read, for several endpoints, built in (no proxy): (www.perplexity.ai page, read 2026-10-22; www.perplexity.ai page, read 2026-10-22; www.perplexity.ai page, read 2026-10-22) None of these pages names a setting for adding a model provider, an endpoint or a base URL. The custom API credentials page covers keys for third-party services Computer calls during a task; it does not describe using one to add a model.
A model on your machine
On Mac, any of 3 local models (a Perplexity model, Qwen3.6 35B-A3B or Gemma 4 E4B) downloads in one click without Ollama or manual runtime setup; the page gives 24 GB unified memory as the minimum. (www.perplexity.ai page, read 2026-10-19)
Perplexity's help page says that with Portable Computer on Windows and Linux, both the model and the agent run on the device, so a whole task can stay local. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says Windows, Linux and DGX Spark systems run only a single local model at once. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Portable Computer help page says Portable Computer puts the model and agent of Computer on hardware you own, so a task executes on the device and not in the cloud. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Personal Computer help page lists the local models by platform: the Perplexity model, Qwen3.6 35B-A3B and Gemma 4 E4B on an Apple silicon Mac; PPLX 27B and Qwen 27B on AMD Ryzen AI Max Windows PCs or a DGX Spark; PPLX 27B on an RTX Windows or Linux PC. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Portable Computer help page lists PPLX 27B and Qwen 27B for a DGX Spark and for supported Windows and Linux PCs. (www.perplexity.ai page, read 2026-10-22)
The product launches with three local models, which are a Perplexity model, Qwen3.6 35B-A3B, and Gemma 4 E4B. (www.perplexity.ai page, read 2026-10-19)
Local and cloud in one session
On a Mac with supported hardware, hybrid compute splits each task: cloud models handle reasoning, web search and planning, and a local model on the Mac handles sensitive files and steps. Local models are for Pro, Max and Enterprise subscribers. (www.perplexity.ai page, read 2026-10-19)
Perplexity's Portable Computer help page says that in a local task you can allow a single cloud step by approving the prompt Portable Computer shows. (www.perplexity.ai page, read 2026-10-22)
Plan or account
Pro, Max, and Enterprise subscribers can access local models. (www.perplexity.ai page, read 2026-10-19)
What a local model must meet
Perplexity's Personal Computer help page lists the hardware that can run local models: an Apple silicon Mac with a 24 GB minimum of unified memory, an AMD Ryzen AI Max processor with a 24 GB minimum of GPU-accessible memory (Windows only), an NVIDIA RTX or RTX PRO card with a 24 GB minimum of video memory (Windows or Linux), and an NVIDIA DGX Spark. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Portable Computer help page says its 24 GB video memory requirement applies to the RTX option, and that a DGX Spark satisfies it by itself. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page names Ubuntu for PCs that have a supported RTX card, and DGX OS for a DGX Spark, as the Linux systems it supports for local models. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Portable Computer help page gives approximate download sizes: 19.6 GB for a first Windows setup with PPLX 27B, 19.9 GB more to add Qwen 27B on Windows, and 38.6 GB for a first DGX Spark setup; it says to leave room beyond those figures. (www.perplexity.ai page, read 2026-10-22)
Not found in the pages we read, for what a local model must meet: (www.perplexity.ai page, read 2026-10-22; www.perplexity.ai page, read 2026-10-22) Neither page says a local model must support tool calling.
Not found in the pages we read, for what a local model must meet: (www.perplexity.ai page, read 2026-10-22; www.perplexity.ai page, read 2026-10-22) Neither page gives a context window size for a local model.
Plan, policy and billing
Perplexity's blog post on Hybrid compute for Mac says Enterprise subscribers' admins can set rules across the organization that decide which information stays on the Mac, which is masked before cloud use and which needs the user's approval first, and that admins can additionally audit each time information leaves the device. (www.perplexity.ai page, read 2026-10-19)
Perplexity's help page says steps the local model handles use no Computer credits, while a cloud step draws on credits the way other Computer work does. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says that for an account an organization manages, local inference is controlled by an administrator: on Windows and Linux the administrator allows or refuses it, and on Mac the administrator can set which material is to remain on the device and which needs your approval before going to the cloud. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Portable Computer help page says Portable Computer is for Pro and Max subscribers, whether their plan is personal or Enterprise, using a DGX Spark or a supported NVIDIA RTX PC on Windows or Linux. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says that on the Free plan the automatic model option and Pro Searches up to a limit are available, and an advanced Search model cannot be picked by hand. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Computer for Enterprise help page says administrators can switch Computer off for the entire organization, at any time. (www.perplexity.ai page, read 2026-10-22)
Perplexity's firewall-policy help page says Enterprise admins can set the websites and network destinations that Computer's cloud sandbox reaches during a task. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Model Council help page says Model Council is included with Max and with Enterprise Max, whether on the web or in the iOS app, and that it is not yet offered as its own search mode in the desktop apps or the Android app. (www.perplexity.ai page, read 2026-10-22)
Perplexity's credits help page says Pro and Max subscribers, including those on Enterprise Pro and Enterprise Max, can get Computer credits. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Portable Computer help page says that on an account an organization manages, an administrator might first have to allow local inference before Portable Computer can be used. (www.perplexity.ai page, read 2026-10-22)
A separate model for planning and for doing
Perplexity's Portable Computer help page says its local model acts as the orchestrator and that a single step can be escalated to frontier models for advanced reasoning, offered a choice of 15+ models. (www.perplexity.ai page, read 2026-10-22)
Works with other products
Perplexity's Portable Computer help page says it can use MCP servers that run on your own machine, only in Local mode, and that cloud Computer does not use servers added under Settings, MCP servers. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Portable Computer help page says it reaches connected apps through its local orchestrator, among them GitHub, Slack, Gmail, Google Drive, Microsoft Outlook, OneDrive and Word. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says local MCP servers are supported in the Mac app, after a helper application named PerplexityXPC is installed. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says you add a remote MCP connector from Plugins in the sidebar by selecting Add custom. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says that when Computer must reach a third-party API with no built-in connector, you can supply your own API credentials. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says a custom API credential has one of three scopes: a single task, your personal credential vault, or a project shared with its team. (www.perplexity.ai page, read 2026-10-22)
About the product
The Personal Computer app for Mac and Windows is rebuilt around Computer, the web-based AI feature from Perplexity. (www.perplexity.ai page, read 2026-10-19)
Perplexity Computer is web-based. It executes tasks in the cloud. (www.perplexity.ai page, read 2026-10-19)
Perplexity's hybrid compute page says that on a Mac, a Computer task begins in the cloud. (www.perplexity.ai page, read 2026-10-19)
The help page says Personal Computer, the desktop app from Perplexity for Mac and Windows, gives Perplexity access to an always-on desktop experience, voice control, native apps and local files, and that installed-app control is not yet available on Windows. (www.perplexity.ai page, read 2026-10-19)
Perplexity's Portable Computer help page says Portable Computer ships inside the Perplexity app for Windows and the Perplexity Linux package, comes with the Pro and Max plans, and has no separate download or subscription. (www.perplexity.ai page, read 2026-10-22)
Perplexity's help page says the Perplexity app for Windows and Mac adds to the web experience: it can work with folders and files held on your computer, the desktop apps and browsers it supports, and Remote Access to that device. (www.perplexity.ai page, read 2026-10-22)
Perplexity's Remote Access help page says desktop work can be begun, resumed and watched from another desktop, from mobile or from the web. (www.perplexity.ai page, read 2026-10-22)
Our reading of vendors' pages, as of the dates shown. Confirm on the vendor's page before relying on any line that cites it. To report an error write to report@crinaro.ai