Summary
Perplexity Portable Computer is now available to subscribers on supported Windows PCs powered by AMD Ryzen AI Max Series processors. AMD says users can run local AI tasks and recurring workflows without using Perplexity Computer credits for inference.
Perplexity’s Portable Computer is now available to subscribers on supported Windows PCs powered by AMD Ryzen AI Max Series processors, including the AMD Ryzen AI Halo developer platform, AMD announced on September 24, 2026. The company says users can run local AI tasks and recurring workflows on their own hardware without using Perplexity Computer credits for inference.
What runs on the PC
Portable Computer brings together AI models, agent tools and controls. AMD says the combination lets users run local AI workflows on supported systems, while retaining the option to use cloud computing when needed. An AI agent is software that can carry out tasks through a series of steps; the announcement describes these systems as working across applications and everyday workflows.
Running inference locally means the PC performs the computation for a task rather than relying on a cloud service for that inference. In this case, AMD says eligible subscribers can do so without drawing on Perplexity Computer credits for inference. The announcement does not say that every task or model runs locally: it also describes cloud use as an option.
Why Ryzen AI Max matters
AMD says Ryzen AI Max processors combine CPU and GPU performance with a large pool of shared memory, allowing local models to run alongside everyday applications. CPUs handle a broad range of computing tasks, while GPUs are designed to perform many operations in parallel; shared memory gives them access to a common pool of data. That combination is intended to support AI workloads on the same PC used for other applications.
The announcement’s practical change is the pairing of Perplexity’s agent software with this class of PC hardware. For subscribers with a supported Windows system, it adds a way to run local tasks and recurring workflows without using inference credits, with cloud computing available when a workflow calls for it.