Perplexity Brings Portable Computer to Windows PCs With 24GB RTX GPUs

What is Perplexity Portable Computer for Windows?

Perplexity Portable Computer for Windows is a local AI agent that runs key parts of its workload on a compatible PC rather than sending every task to a cloud service. Perplexity made the feature available through its existing Windows app on September 14, 2026, extending support beyond NVIDIA DGX Spark and Linux systems. Perplexity’s product announcement confirms the Windows release and subscription availability.

Perplexity Portable Computer runs a local model, agent harness, orchestrator, and scheduler on the device. The agent harness is the software layer that carries out multi-step work, while the orchestrator and scheduler coordinate the sequence of tasks. This matters because work involving local files and tools can remain off the cloud when the local system handles the task.

Perplexity Portable Computer is not a replacement for every online AI feature. Perplexity allows users to authorize a cloud-model escalation when a task needs current information or more advanced reasoning. The practical distinction is straightforward: local operation prioritizes on-device processing, while cloud escalation adds broader model access when the user decides that tradeoff is appropriate.

Which Windows PCs can run Perplexity Portable Computer?

Perplexity Portable Computer for Windows requires an NVIDIA GeForce RTX or RTX PRO GPU with at least 24GB of VRAM. VRAM is the graphics card’s dedicated memory, and local AI models need substantial VRAM because model data must be loaded and processed close to the GPU. A Windows PC with an RTX-branded GPU does not automatically qualify if it has less than 24GB of VRAM.

NVIDIA says the simplified setup can use Qwen 3.8 27B, a model post-trained for Perplexity Computer and optimized for RTX GPUs. NVIDIA’s launch details also list the supported connector ecosystem and describe the local Windows deployment. The stated hardware requirement means Portable Computer is aimed at high-end desktops, workstations, and similarly equipped systems rather than typical mainstream laptops.

Perplexity Portable Computer’s 24GB threshold is the first check to make before considering a subscription change or workflow migration. Independent hardware coverage also identifies the same RTX and VRAM requirement, reinforcing that the limitation is central to the Windows launch rather than a minor configuration detail. Hardware coverage of the Windows rollout provides additional context on the unusually high GPU-memory requirement.

How does local processing affect privacy and sensitive files?

Perplexity Portable Computer can keep work across sensitive files and local tools off the cloud because the model and agent software run on the Windows device. This is the primary privacy benefit of the feature. A local workflow can reduce the need to transmit file contents to an external service when the task can be completed on the PC.

Local processing does not mean every action is automatically private or isolated from online services. Portable Computer supports connectors for Outlook, OneDrive, Word, Google Drive, Gmail, Slack, and GitHub, and those services have their own account permissions, sharing rules, and retention policies. Users should review what each connected account can access before allowing an agent to work across documents, messages, or code repositories.

Perplexity Portable Computer also offers cloud escalation, which changes the data-handling decision for the specific task that uses it. The most sensible approach is to keep confidential files and internal work in local workflows when possible, then authorize a cloud model only after deciding that the need for current information or more advanced reasoning outweighs the additional data-sharing consideration.

What tasks and connectors does Portable Computer support?

Perplexity Portable Computer supports connectors for Outlook, OneDrive, Word, Google Drive, Gmail, Slack, and GitHub, according to NVIDIA. These connectors matter because an agent becomes more useful when it can work across the services where documents, messages, files, and development projects already reside. The supported list covers both Microsoft and Google productivity environments, plus workplace communication and code hosting.

Perplexity said on September 15, 2026, that its Windows app can orchestrate more than 15 cloud models. Perplexity also said that the app can be used in Excel, Word, PowerPoint, Outlook, and Teams on supported HP hardware. That application-level access suggests a workflow focused on productivity tasks that span office documents and communications, rather than a standalone chatbot window.

Windows users should still treat connected-agent access as a permission decision, not a default productivity setting. A task that reads email, searches cloud storage, or interacts with GitHub can touch valuable information even when the core model runs locally. Users who already rely on cross-app workspace tools should apply the same access discipline: connect only the accounts needed for the task and review permissions regularly.

When does Perplexity Portable Computer use the cloud?

Perplexity Portable Computer uses local processing by default for supported on-device work, while cloud models are available when the user authorizes an escalation. Perplexity says users can make that choice for current information or more advanced reasoning. The feature therefore gives the user a clearer decision point between local execution and cloud-assisted work.

Current information is a meaningful limitation for a local model because news, product changes, live availability, and other time-sensitive material can change after the model’s training or local preparation. More advanced reasoning is also a separate requirement from privacy. A user may prefer local work for confidential drafts or files, but choose a cloud model for a task that needs broader model capability or fresh information.

Perplexity says local work does not consume Perplexity Computer credits. That distinction can make local use more attractive for repeated work on eligible hardware, although the hardware requirement remains substantial. Users should not assume that a locally completed task has the same capabilities as a cloud-assisted one, because Perplexity presents cloud escalation as the option for tasks that need additional information or reasoning capacity.

Who can access Perplexity Portable Computer on Windows?

Perplexity Portable Computer for Windows is available to Perplexity Pro, Max, and Enterprise subscribers through the existing Windows app. The rollout therefore has 2 access requirements: an eligible subscription and a compatible NVIDIA RTX or RTX PRO GPU with at least 24GB of VRAM. Meeting only one requirement is not enough for on-device inference.

Perplexity’s Windows launch also means the feature is not limited to dedicated AI hardware or Linux systems. The expansion is important for organizations and advanced individual users that already work in Windows productivity software but want the option to keep some AI tasks local. At the same time, the VRAM requirement restricts the immediate audience to systems built for demanding graphics, AI, or workstation workloads.

Perplexity Portable Computer does not change the broader competitive pressure around desktop AI tools, where cloud-based assistants remain easier to access on ordinary PCs. Google, for example, recently introduced a dedicated Windows experience through its Gemini Windows app, but local processing requirements and privacy controls differ by product. Windows users should compare the operating model, account permissions, and hardware needs rather than treating all desktop AI assistants as interchangeable.

How should Windows users set up Portable Computer safely?

Perplexity Portable Computer should be set up only after confirming the PC has the required NVIDIA GPU and at least 24GB of VRAM. The hardware check matters because local inference depends on the GPU memory threshold, and an unsupported system cannot provide the intended on-device experience. Open Windows system information or your NVIDIA hardware utility to verify the installed graphics card before attempting to use the feature.

1. Confirm that your Windows PC has an NVIDIA GeForce RTX or RTX PRO GPU with at least 24GB of VRAM.
2. Open the existing Perplexity Windows app with a Pro, Max, or Enterprise subscription.
3. Review the local-processing option and understand when a task will remain on the device.
4. Connect only the accounts and services required for the work, such as OneDrive, Gmail, Slack, or GitHub.
5. Authorize cloud escalation only for tasks that need current information or more advanced reasoning.

Perplexity Portable Computer users should also avoid giving an agent broad access to personal or workplace accounts without reviewing the practical consequences. Connected cloud services can contain contracts, customer data, financial documents, private messages, and source code. Account compromises can also create exposure beyond the AI tool itself, as incidents involving fraudulent data requests have shown in other services.

Stop and contact Perplexity support, your employer’s IT administrator, or the PC maker’s support team if the app requests unexpected permissions, fails during a hardware-dependent setup, or behaves differently from the documented local workflow. A normal user should not attempt unsupported GPU modifications or system-level changes to force local AI inference on hardware that does not meet the requirement.

Is Perplexity Portable Computer worth using on Windows?

Perplexity Portable Computer is worth considering for Windows users who already have the required 24GB RTX or RTX PRO hardware and want local handling for sensitive files or tool-based workflows. The feature’s clearest benefit is that its local model, agent harness, orchestrator, and scheduler can run on the device. That design can reduce cloud exposure for tasks that do not require outside information.

Perplexity Portable Computer is not a practical purchase justification for most Windows users because the launch requires specialized GPU hardware and a paid Perplexity subscription. The feature is more relevant to people who already own eligible systems, including high-end desktop and workstation users, than to someone using a standard office laptop. Cloud AI tools remain the more accessible choice for users without the required hardware.

For qualified users, the practical decision is to identify which workflows benefit from local processing and which require cloud assistance. Local document and file work may fit the feature’s privacy model, while research involving current events or changing facts may justify a cloud escalation. The strongest use case is therefore a mixed workflow that keeps appropriate work local without assuming local AI can replace every online capability.

FAQ

Does Perplexity Portable Computer work on every Windows PC?

Perplexity Portable Computer does not work on every Windows PC because on-device inference requires an NVIDIA GeForce RTX or RTX PRO GPU with at least 24GB of VRAM. The feature also requires a Perplexity Pro, Max, or Enterprise subscription.

Does Perplexity Portable Computer send local files to the cloud?

Perplexity Portable Computer can keep supported work involving sensitive files and local tools off the cloud because its model and agent software run on the Windows device. Cloud processing can still occur when a user authorizes an escalation for current information or more advanced reasoning.

Do local Portable Computer tasks use Perplexity Computer credits?

Local Portable Computer tasks do not consume Perplexity Computer credits, according to Perplexity. Cloud escalation is a separate option that users authorize when a task needs current information or more advanced reasoning.

Which apps can Perplexity Portable Computer connect to?

Perplexity Portable Computer supports Outlook, OneDrive, Word, Google Drive, Gmail, Slack, and GitHub connectors, according to NVIDIA. Users should connect only the accounts needed for a task because each connected service can contain sensitive information.

Can Perplexity Portable Computer use cloud models?

Perplexity Portable Computer can use cloud models after the user authorizes an escalation. Perplexity says cloud escalation is available for current information or more advanced reasoning that the local workflow does not provide.