Perplexity’s Portable Computer moves its agentic workload onto a user’s machine so models, files, and work stay local. The trade-off: faster, private, and cheaper runs for paid users who can meet demanding GPU and OS requirements.
Portable Computer runs Perplexity’s agent locally, keeping models and sensitive data on your device and reducing cloud inference costs.
Local execution delivers faster responses and stronger isolation, but Portable Computer is limited to paid accounts and requires high-end NVIDIA GPUs and Linux (now) or Windows soon.
Perplexity has introduced Portable Computer, a local-first version of its agentic Personal Computer. The agent executes AI tasks on the user's machine, running models, handling files, and completing assignments inside an isolated sandbox. Users control if and when the agent needs cloud access to finish tasks or fetch external data.
Three concrete advantages of local execution are performance, security, and cost. Local models can respond faster than cloud-hosted frontier models because inference happens on-device. Data processed by the agent stays on your machine unless you explicitly permit cloud access, reducing the surface for data exposure. And local inference does not consume cloud credits or tokens, so only actions that require going online will incur billing costs.
Local AI is resource-intensive and Portable Computer reflects that. On Linux, Perplexity requires an NVIDIA DGX Spark or other machine with an NVIDIA RTX GPU, running Nvidia DGX OS or Ubuntu on ARM or x64. Windows support will demand an NVIDIA RTX GPU with at least 24GB of VRAM. Lower-end desktop GPUs (for example, those with 4GB VRAM) will not meet the threshold. High-VRAM RTX cards carry significant cost, so adopting Portable Computer may require new hardware purchases for many users.
Portable Computer supports at least two model options. Qwen 3.8 27B is an open-weight model noted for speed and capability across coding, research, and other complex agentic tasks. Perplexity also offers PPLX 27B, a post-trained variant the company applies to Portable Computer to sharpen accuracy and efficiency for agent workloads. Perplexity plans to add Nvidia's Nemotron 3.5 Lightning for heavy-volume or long-running tasks. Users can switch models based on the assignment and their hardware constraints.
How it interacts with cloud services Even when running locally, Portable Computer can access cloud services such as Google Drive, Gmail, Slack, and GitHub to pull in remote files or accounts when the user permits. Perplexity's model routes allow the agent to keep sensitive data local while still using cloud resources where needed, with only the cloud portions consuming credits.
Agentic assistants have a structural problem: the context that makes them useful — deal documents, privileged files, client records — is exactly the context users cannot send to a cloud endpoint. This week, Perplexity shipped its answer for Mac. Hybrid compute splits a single Perplexity Computer task between frontier m
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.
Your gaming GPU can now moonlight as an AI agent.
Perplexity has launched Hybrid Compute for its Computer agent on Apple Silicon Macs, a system that splits a single task between frontier… The post Perplexity brings hybrid compute to Mac, keeping sensitive AI work on device appeared first on MacDailyNews .
Aside from being featured by Apple in the M6 Mac mini launch, Perplexity has been relatively quiet on the Mac front this summer. That changes today with hybrid compute for Mac. more…
The AI pendulum has swung back and forth throughout the last few years, from ChatGPT and Copilot in the cloud, then back to local AI and the PC’s newfound NPU, then back to independent agents like Claude. Now it’s headed the way of the PC once again, abandoning privacy arguments (which still stand strong!) for a more s
Loading more related stories...
Open the app view to save this story, compare related coverage, and continue from the same source.