Perplexity Portable Computer became available on Linux for NVIDIA RTX GPUs with 24GB or more of VRAM
Following·since 3 Sep 2026·New·2 sources
NVIDIA RTX Spark ships October 2026; PAIR and AI agent tools announced
The gist
RTX Spark brings a high-memory AI-focused compact PC to the consumer market in October 2026. The accompanying software stack, including PAIR for multi-PC inference pooling, llama.cpp and vLLM optimizations, and simplified agent setup, reduces the friction of running AI workloads entirely on local hardware.
The full picture
RTX Spark compact PCs are confirmed to ship in October 2026, with Lenovo's Yoga Pro 9n, Yoga 9n 2-in-1, and an Acer compact desktop concept shown at IFA 2026. The hardware features a 1 Petaflop RTX Blackwell GPU, up to 128GB of unified memory, and a 20-core Grace CPU. Alongside the hardware, NVIDIA announced PAIR (Personal AI Router), a free open-source tool that routes AI inference requests across multiple PCs on a local network, adapting dynamically as devices join or leave. NVIDIA also contributed kernel optimizations and speculative decoding improvements to llama.cpp and vLLM, delivering up to 1.9x throughput gains on an RTX 5090. Three AI agent applications, Hermes Agent, OpenClaw, and Perplexity Portable Computer, are adding simplified or one-click local model setup on NVIDIA hardware. Perplexity Portable Computer, which runs its orchestrator LLM, subagent LLM, and agent harness entirely on local hardware with no cloud dependency, is now available on Linux for NVIDIA RTX GPUs with 24GB or more of VRAM. DLSS 4.5, featuring a second-generation transformer model for Ray Reconstruction, was unveiled at Gamescom. MLCommons released MLPerf Client v2.0 with new agentic AI and image generation benchmarks, developed with NVIDIA.
How it developed
Sources
The daily email
Want this in your inbox?
I send a short email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free