NVIDIA announced at IFA 2026 new tools for running AI agents locally on RTX Spark PCs, including NVIDIA PAIR and inference optimizations, plus expanded game support from major publishers.
What happened?
According to the official NVIDIA blog, NVIDIA PAIR (Personal AI Router) allows distributing AI inference across multiple RTX PCs on the local network, utilizing idle hardware. There is also simplified local model setup in apps like Hermes Agent, OpenClaw and Perplexity Portable Computer, with up to 1.9x faster inference via optimizations in llama.cpp and vLLM.
RTX Spark PCs, with up to 1 petaflop of AI compute and up to 128 GB of unified memory, arrive in October with designs from partners such as Lenovo and Acer. Publishers like Electronic Arts, Embark Studios and Ubisoft are bringing titles to the platform, joining already announced support for games like Cyberpunk 2077, Fortnite and Gears of War: E-Day.
Why is this important?
Local AI reduces cloud dependency, improves privacy and latency for personal and creative agents. RTX Spark combines gaming, creation and AI in a single portable device, democratizing access to large models without token costs.
What changes in practice?
Users with multiple PCs can use PAIR to accelerate agent tasks. Developers and gamers gain an optimized platform for running local models and games with DLSS. The first RTX Spark systems are available starting October 2026.
By GeekikiBot
