NVIDIA RTX Spark Launches Globally, Focusing on Local LLM
On Oct 16, NVIDIA held an RTX Spark launch experience event at JD MALL in Beijing's Shuangjing area from 7 p.m. to 9:30 p.m., drawing fans and KOLs to try on-device LLM demos and play 3A games on thin-and-light laptops. The event was packed. At 8 p.m. the same day, the RTX Spark officially went on sale worldwide.
Positioned as the world's first Windows PC chip designed for personal AI agents, RTX Spark integrates Blackwell RTX graphics, CUDA, and AI capabilities into a single chip, with up to 128GB of unified memory. It can run LLMs with over 100 billion parameters locally without relying on cloud computing.
The first laptops equipped with RTX Spark come from Lenovo, ASUS, MSI, Dell, and Microsoft, in two tiers: a high-end 6144-core GPU version and a mainstream 5120-core GPU version. The ASUS ProArt P16 RTX Spark 128GB model sold out quickly at its 37,999 yuan launch price. Prices start at 18,999 yuan, with up to 1,500 yuan off via national subsidies.
On the software side, Adobe is adding TensorRT acceleration to Photoshop, Premiere, and other tools, while CapCut PC, Bilibili Live, and D5 Lite support RTX Spark on day one. The platform runs AI agents like GitHub Copilot and Perplexity, and supports hundreds of games. Jensen Huang said RTX Spark combines NVIDIA AI, RTX graphics, and CUDA with Windows in one PC.
