Nvidia Unveils 'RTX Spark' AI Notebook at COEX with Up to 128GB Memory
Nvidia has unveiled the RTX Spark AI notebook combining Blackwell GPU and Grace CPU at COEX in Seoul. Discover the token-free local AI agent environment, core s
Nvidia has officially unveiled the "RTX Spark" AI notebook tailored for next-generation local AI environments at an event held at COEX in Seoul. The new device is engineered to provide a local AI agent environment running directly on the device without cloud dependency, allowing users to leverage advanced AI models without per-token charges.
Image source: Nvidia
The RTX Spark is equipped with a superchip combining Nvidia's latest Blackwell GPU and Grace CPU. The top-tier configuration offers up to 128GB of unified memory and delivers up to 1 PFLOPS of AI compute performance based on FP4. This hardware configuration enables seamless execution of complex office task automation, video editing, and programming assistance directly in a local environment.
During the launch event, Nvidia demonstrated various practical AI agent workflows targeted at developers and creators, including automated email prioritization, workspace organization via Google Calendar and Drive integration, and automated video clip extraction and vertical reformatting. Nvidia announced that the top-spec model targeting developer and creator markets is priced at 9 million KRW.
Nvidia plans to aggressively target the local developer and creator ecosystems by providing a secure, cloud-independent local AI notebook experience. Further software optimizations and ecosystem expansions are expected to maximize on-device AI performance.