Nvidia launches RTX Spark laptops designed to run AI agents locally

What happened

Nvidia announced the launch of RTX Spark laptops on Friday, October 16, 2026. These devices are built to run AI agents locally on the hardware rather than relying on the cloud. The laptops use a single superchip designed with help from MediaTek, which combines a Blackwell RTX graphics chip with a Grace processor [1].

Key technical specifications and software details include:

• Hardware Performance: The laptops offer up to 128GB of shared memory and up to 1 petaflop of AI performance [1].
• AI Models: Nvidia released the Nemotron 3.5 Bolt model on Thursday. It has 74 billion parameters, with 9 billion active at a time, and fits into 24GB of memory [1]. A smaller 4GB version for voice assistants and game characters is scheduled for release on 23 October [1].
• Software Compatibility: Nemotron 3.5 Bolt works with llama.cpp, Ollama, LM Studio, Unsloth, and vLLM [1]. Adobe has optimized Photoshop, Premiere, and Substance 3D for the chip; Photoshop and Premiere are currently in beta, while Substance 3D is expected soon [1].
• Gaming: Hundreds of games are available at launch, including Cyberpunk 2077, Fortnite, and NBA 2K27. As of Friday, NBA 2K27 features a DLSS 5 model that can increase frame rates by up to 1.4 times [1].
• Developer Tools: Nvidia and Microsoft released the first PyTorch build with CUDA for Windows on Arm, noting that 97% of the top 15,000 Python packages are ready [1].
• Pricing and Models: Prices start at $2,599 [1]. Available models include the ASUS ProArt P14 and P16, Dell XPS 16, HP OmniBook Ultra and OmniBook X 14, Lenovo Yoga Pro 9n, Microsoft Surface Laptop Ultra, and MSI Prestige N16 and N16 Flip AI+ [1].

The Microsoft Surface Laptop Ultra is a flagship model in this new fleet [2]. It features a 15-inch Mini-LED touchscreen with a 120-Hz refresh rate and a pixel density of 257 pixels per inch [2]. It includes ports such as HDMI, USB-A, an SD card slot, and three USB4 Type-C ports, one of which has a magnetic feature [2].

Who is affected

• Content Creators and Gamers: Users can run personal AI agents, play high-end games, and use optimized Adobe software [1].
• Developers: Those using Python or PyTorch can utilize the new Windows on Arm support [1].
• General Users: People can use the free Nvidia PAIR tool to serve AI models to other computers on a home network running Windows, Linux, or macOS [1].

Nvidia founder and CEO Jensen Huang stated the laptops bring together Nvidia AI, RTX graphics, and CUDA with Microsoft Windows [1]. Microsoft executive vice president Yusuf Mehdi said the goal is to make Windows a platform for hybrid intelligence [1]. Dillon Rolnick, CEO of Nous Research, noted that the inclusion of Hermes allows users to work with open AI models locally with more data control [1].