Slim, lightweight NVIDIA RTX Spark Windows laptops designed for every task are getting new models, agentic tools, and games.
Windows RTX Spark laptops gain new models, Nemotron 3.5 Transcribe, GPU sharing, and Liquid AI support for local AI, creative, gaming.
Key Takeaways
- NVIDIA RTX Spark laptops are shipping from major PC makers, combining a Blackwell RTX GPU and Grace CPU in one superchip with up to 128GB unified memory, targeting premium Windows AI, creative, and gaming workloads.
- The unified-memory design lets AI models and demanding apps run together locally, enabling desktop-class creative and AI tasks on thin, light laptops with all-day battery life.
- NVIDIA introduced Nemotron 3.5 Transcribe, an open-weight speech model for transcription, speaker attribution, translation, and audio Q&A across 40 languages, with local deployment options for RTX Spark and other NVIDIA devices.
- New RTX Spark tools include a GPU resource-sharing control for running AI agents alongside apps, day-zero support for Liquid AI’s LFM3 models, and game enhancements such as Phantom Blade Zero with DLSS 4.5 and RTX Remix 1.6 for path-traced Minecraft.
NVIDIA's RTX Spark laptops are now shipping from six PC makers, pairing a Blackwell RTX GPU and a Grace CPU on a single superchip with up to 128GB of unified memory and up to 1 petaflop of FP4 AI performance. The machines start at $2,599, with 128GB configurations already on shelves. ASUS, Dell, HP, Lenovo, Microsoft and MSI are all offering models, and NVIDIA is positioning them as premium Windows laptops that put the CUDA platform, RTX graphics and personal AI supercomputer capabilities in a portable chassis.
This matters beyond the spec sheet. The same memory pool feeds both creative applications and AI models, so a laptop can handle work that previously required a desktop workstation, even on battery. NVIDIA says the systems are as thin as 14mm and as light as three pounds, with all-day battery life. Buyers can choose among different screen sizes, colors, touchscreens, pen support and two-in-one convertibles.
What the Hardware Means for Three Kinds of Buyers
For AI developers, RTX Spark runs models and agents locally on CUDA, the same platform behind cloud AI, with a deployment path across more than 150 million NVIDIA RTX AI PCs and NVIDIA GPUs in the cloud. Creators get the ability to edit camera-native 4:2:2 video and render 3D scenes of up to 90GB; Adobe Photoshop and Premiere offer native RTX Spark support through beta releases. Gamers can run hundreds of titles with DLSS, ray tracing, Reflex and G-SYNC.
The thin design is possible because the GPU, CPU and unified memory sit on one power-efficient superchip rather than separate components with their own memory pools.
Nemotron 3.5 Transcribe Turns Recorded Audio Into Answers
NVIDIA introduced Nemotron 3.5 Transcribe, an open weight speech model that handles transcription, speaker attribution and translation in a single promptable model, plus questions about the audio. It covers 40 languages, including seven Indic languages, and developers steer it with plain-language prompts. For a single recorded meeting, for example, a user can ask it to transcribe the recording and label each speaker, translate the audio into English, or list the decisions made in the meeting. Prompts can also carry product terms and formatting conventions, which helps with specialized vocabulary in healthcare, financial services and customer service.
Because the weights are open, developers choose where audio is processed, from a desktop to the cloud. NVIDIA says the model ranks No. 1 on the Hugging Face Open ASR and No. 2 among open weight models on Artificial Analysis, with approximately 4.3 billion active parameters—80% fewer than the No. 1 open weight model. It is available today on Hugging Face. Quantized versions are coming soon for local use on RTX Spark, DGX Spark and Jetson AGX Thor devices, or scaled up on Blackwell data center GPUs; NVFP4 versions will be deployable through NeMo-Speech.cpp or vLLM.
A New Dial for Sharing the GPU Between Apps and Agents
RTX Spark users can run long-running agents alongside their main applications, and both draw on the same GPU compute. A new resource management feature in NVIDIA App lets users decide how that compute is split: prioritize the foreground app for best performance, or give a background agent more resources to finish work faster. The control lives under System > Advanced in the NVIDIA App and requires hardware-accelerated GPU scheduling to be enabled in Windows Settings.
Liquid AI's LFM3 Family Lands With Day-Zero Support
Liquid AI's LFM3 is a new generation of Liquid Foundation Models built for specialized tasks, spanning text, vision, audio, omni and decision models on one shared backbone, with base and post-trained versions available as open weights. The models are compact and fast: Liquid AI's new 1.7-billion-parameter base model outperforms its previous 2.6-billion-parameter base model. They support local workflows such as agents running in Hermes Agent, Pi and OpenClaw to triage inboxes, manage calendars, read documents and app screens, transcribe and translate speech, answer spoken questions about handwritten math, and route support tickets or voice requests. Developers can run LFM3 on RTX Spark through native support in llama.cpp and vLLM, and fine-tune the models locally.
Phantom Blade Zero Launches With Ray Tracing and DLSS 4.5
S-GAME's Phantom Blade Zero launches tonight, Oct. 28, at 7 p.m. PT on Steam and the Epic Games Store. Players take the role of Soul, a swordsman framed for his master's murder, traveling the martial world to uncover a conspiracy. NVIDIA worked with S-GAME Studio and S-GAME Publishing to add ray-traced effects: more accurate bounced lighting, detailed reflections and realistic shadows on foliage and terrain. DLSS Dynamic Multi Frame Generation boosts frame rates with up to 6x frame generation, or adjusts the multiplier on the fly to hit a player's target frame rate.
RTX Remix 1.6 Brings Path Tracing to Minecraft
RTX Remix, NVIDIA's modding platform for remastering classic games with path tracing and DLSS, released version 1.6 last week, bringing the toolset to Minecraft. Creators and players can render Minecraft Java Edition 1.21.1 through RTX Remix with DLSS-accelerated path tracing and use realistic lighting and materials in future creations. The update adds DLSS 5 with 3D-Guided Neural Rendering, up to 6x Multi Frame Generation, native RTX Spark support and agent-assisted workflows. Starting today, those upgrades come to Portal with RTX and the Half-Life 2 RTX demo, both now supporting RTX Spark with DLSS 3D-Guided Neural Rendering, a new Performance Mode for DLSS Frame Generation and options that speed up path tracing.
AI-powered modding also gets simpler. A redesigned ComfyUI workflow lets modders generate physically based materials directly inside RTX Remix with no API setup, batch processing for full projects and presets for less technical creators. With NVIDIA agent skills, compatible AI agents can work directly with RTX Remix to speed up remastering projects.
More From NVIDIA's Local AI Push
During the GTC Berlin keynote, NVIDIA founder and CEO Jensen Huang showed how AI agents running on RTX Spark can help train Hugging Face's Microduck robot. A technical blog walks through the workflow: Hermes Agent and Nemotron 3.5 Bolt plan an obstacle course on RTX Spark, Astra builds the 3D assets, NVIDIA Isaac Lab trains the navigation policy and RTX Spark validates it in NVIDIA Isaac Sim.
NVIDIA Nemotron 3.5 Bolt E4B 4GB is now available, enabling natural conversations and multistep agent tasks with as little as 4GB of memory; it powers local voice assistants, game characters and embedded interfaces on RTX Spark, DGX Spark, GeForce RTX GPUs and NVIDIA Jetson devices, and supports llama.cpp, vLLM, Ollama, Unsloth and LM Studio, working with Hermes and OpenClaw. For heavier agent workloads, Nemotron 3.5 Bolt 74B-A9B 24GB is designed to run local agents on RTX Spark.
Black Forest Labs' FLUX 3 Image Klein is a 5.6-billion-parameter open weights model for generating and editing images locally. NVIDIA-optimized versions run on RTX GPUs with as little as 8GB of VRAM, and developers can fine-tune it on RTX Spark using an NVIDIA NeMo AutoModel recipe to customize it for specific creative workflows.
NVIDIA DGX Spark 64GB is now available from manufacturer partners, powered by the GB10 Grace Blackwell Superchip with 64GB of unified memory, DGX OS and the NVIDIA AI software stack for agent development, inference and fine-tuning.
What to Watch
The RTX Spark launch is really a bet that local AI work—transcription, agent orchestration, image generation, game modding—belongs on a laptop rather than a cloud account. The pieces are now in place: open weight models small enough to run locally, software controls to divide GPU resources between people and agents, and hardware that shares one memory pool across CPU and GPU tasks. Whether buyers treat a $2,599 laptop as a workstation replacement will depend on how quickly the promised quantized models and beta Adobe support mature. NVIDIA says 128GB configurations are already on shelves, so the answer will show up in real workflows soon.

Comments (0)