Nvidia RTX Spark Laptops Add Transcription Model and New Game Support

Nvidia used an October 28 blog post to lay out a round of updates for its RTX Spark line of Windows laptops, the thin and light machines built around a combined Blackwell GPU and Grace CPU superchip. The laptops, sold by Asus, Dell, HP, Lenovo, Microsoft and MSI, are aimed at developers, creators and gamers who want to run AI workloads locally rather than through a cloud subscription, and the company framed this update as widening what that local hardware can do.

The most notable addition is Nemotron 3.5 Transcribe, an open weight speech model that handles transcription, speaker labelling and translation from a single promptable system rather than three separate tools. It covers 40 languages, including seven Indic languages, and developers can steer it with plain language instructions such as asking it to list decisions made in a meeting. Nvidia says the model ranks first on the Hugging Face Open ASR leaderboard and second among open weight models on Artificial Analysis, while running with far fewer active parameters than the leading open weight alternative.

Alongside the speech model, RTX Spark laptops gain day-zero support for Liquid AI’s LFM3 family, a set of compact models spanning text, vision, audio and decision tasks built on a shared backbone. Nvidia also added a resource management control inside Nvidia App that lets a user decide whether a foreground application or a background running agent gets priority access to GPU capacity, a setting that requires hardware accelerated GPU scheduling to be turned on in Windows first.

On the gaming side, S-GAME’s Phantom Blade Zero launched the same day with ray tracing and DLSS support, and RTX Remix version 1.6 brought path traced rendering to Minecraft Java Edition along with new support for RTX Spark hardware itself.

The numbers

Nvidia lists RTX Spark laptops starting at $2,599, with configurations carrying up to 128GB of unified memory shared between the GPU and CPU rather than split into separate pools. That shared memory design is what lets the company claim a laptop can take on tasks that previously needed a desktop workstation. Nemotron 3.5 Transcribe is described as running with roughly 80% fewer active parameters than the top open weight model it is compared against, a reduction Nvidia ties directly to making transcription practical on local hardware rather than a data centre. In gaming, Phantom Blade Zero supports up to 6x frame generation through DLSS, a multiplier the system can also adjust automatically to hit a target frame rate.