Logo
FrontierNews.ai

Perplexity's Local AI Agent Now Runs on AMD PCs, Expanding Beyond Nvidia Hardware

Perplexity has expanded its Portable Computer platform, a local version of its AI agent software, to Windows PCs powered by AMD Ryzen AI Max processors. The announcement, made on September 24, 2026, marks a significant step in making on-device artificial intelligence accessible across different hardware platforms. Previously available only on Nvidia systems, Portable Computer now runs on AMD's latest processors, including the Ryzen AI Halo developer platform, giving users another hardware option for running AI agents entirely on their machines without relying on cloud services.

What Makes Running AI Locally on Your PC Different?

Running a capable language model on your personal computer instead of sending requests to a cloud server fundamentally changes how AI works for you. When you use Portable Computer, your files, prompts, and model outputs stay on your device. You maintain control over what information leaves your machine and when. The system bundles together a 27-billion-parameter language model, an agent orchestrator, a scheduler, a sandbox for safe code execution, and a sensitive-content classifier into one integrated package.

The practical advantage is straightforward: local inference does not consume Perplexity's cloud credits, and cloud escalation requires explicit user permission. For organizations handling sensitive data, this matters enormously. Perplexity's target use cases include deal documents, patient records, unreleased code, and payroll exports, exactly the files that compliance teams want to keep off public cloud servers.

Why Does AMD's Hardware Architecture Matter for Local AI?

AMD's Ryzen AI Max processors use a different architectural approach than traditional gaming graphics cards. Instead of a GPU with its own dedicated memory, these processors feature a shared memory pool that both the CPU and GPU can access. This design choice has a direct impact on how much computing power is available for running AI models locally. A qualifying system needs at least 24 gigabytes of GPU-accessible memory, but on a Ryzen AI Max system, the amount available for models scales with the system's overall configuration.

To illustrate the difference, consider HP's ZBook Ultra G3a 16, a mobile workstation that comes with Perplexity Computer preloaded. This system offers up to 192 gigabytes of unified memory with up to 160 gigabytes allocable to graphics. That is far more headroom than a 27-billion-parameter model requires, leaving room for multi-agent workflows and larger models in the future.

"We're entering a new era of computing, where personal AI transforms the PC from a tool into an intelligent partner. We're creating a new kind of agentic PC that can understand, reason and act alongside us," said Jack Huynh, senior vice president and general manager of the Computing and Graphics Group at AMD.

Jack Huynh, Senior Vice President and General Manager of the Computing and Graphics Group at AMD

How to Set Up and Use Portable Computer on Your AMD PC

  • Download and Install: Users download a model in one click from the Local Inference section of the Perplexity app's settings for Windows, and the inference engine and classifier install automatically alongside it.
  • Configure Permissions: Users then connect their apps and choose which folders Portable Computer is allowed to work in, maintaining granular control over file access.
  • Choose Your Model: Two models are available at launch: Qwen 3.8 27B and PPLX 27B, which Perplexity describes as its own post-trained model optimized for local performance.
  • Set Up Integrations: The local orchestrator connects to Gmail, Outlook, Slack, and GitHub, and local Model Context Protocol servers let it work with supported applications installed on the same PC.
  • Schedule Recurring Tasks: Beyond one-off requests, Portable Computer can schedule recurring jobs that run locally without consuming cloud credits, though scheduled tasks run only while the app is open and the PC is awake.

How Does Portable Computer Balance Local Processing with Cloud Power?

Portable Computer is not designed to be completely isolated from the internet. Instead, it uses a hybrid approach where routine tasks and sensitive work stay local, while heavier reasoning and current information come from the cloud when needed. If a task requires information beyond what the local model can handle, the system asks for permission before sending anything to the cloud.

Perplexity's example illustrates this hybrid model in action: a financial analyst working on a local earnings model can pull in current market data from the web through Perplexity Search and use a frontier model to test how changing interest rates might affect a forecast. The local model handles the foundational work, while cloud models handle the complex reasoning.

Security is built into the workflow. An on-device classifier scans content for sensitive details such as names, account numbers, and credentials before anything leaves the PC. If it flags something sensitive, the user can share it anyway, keep that step on the local model, or stop the task entirely. For organizations, administrators on managed accounts decide whether local inference is allowed, giving IT teams control over which teams can use on-device processing.

Why Does Hardware Choice Matter for On-Device AI?

One month after Portable Computer launched on Nvidia hardware, its availability on AMD systems signals an important shift in the on-device AI market. Local agents should not be tied to a single silicon vendor, and more hardware choice benefits users by preventing vendor lock-in and encouraging competition. AMD's announcement represents the second major processor manufacturer to support Perplexity's local agent platform, suggesting that on-device AI is moving beyond a niche feature into a mainstream computing capability.

For AMD, this collaboration extends its broader vision of agentic PCs. The company introduced the first AI PCs in 2023 with the first dedicated neural processing unit in an x86 processor, and followed in 2025 with agentic PCs capable of running AI agents locally. The Perplexity partnership brings these capabilities into the applications and workflows people use every day.

Portable Computer is available now to Perplexity Pro and Max subscribers on individual and enterprise plans through the Perplexity app for Windows on systems with at least 24 gigabytes of GPU-accessible memory. HP's ZBook Ultra G3a is expected to be available in October 2026 with Perplexity Computer preloaded.