What does NVIDIA RTX Spark bring to thin laptops in late 2026?

Hardware and availability

RTX Spark laptops are now sold by ASUS, Dell, HP, Lenovo, Microsoft and MSI. Each system merges a Blackwell RTX GPU with a Grace CPU on one superchip. The package delivers up to 128GB of unified memory and up to 1 petaflop of FP4 AI compute. Prices start at $2,599. Configurations with the full 128GB are already on shelves. The chassis can be as thin as 14mm and as light as three pounds. Battery life lasts all day. Creative projects and AI models draw from the same memory pool. A laptop can handle work that used to require a desktop workstation even when unplugged.

New AI models and local tools

Nemotron 3.5 Transcribe arrives today as an open weight speech model. It handles transcription and speaker attribution in one promptable model; translation comes as a third capability. The model covers 40 languages including seven Indic languages. Developers guide it with plain language prompts. For a single recorded meeting users can ask it to transcribe and label each speaker. They can ask for translation into English. They can ask for a list of decisions made. Prompts may include product terms and formatting conventions. This helps with specialized vocabulary in healthcare and financial services; customer service is another field mentioned. Because the weights are open developers choose where audio is processed from a desktop to the cloud. Nemotron 3.5 Transcribe ranks number one on the Hugging Face Open ASR leaderboard. It ranks number two among open weight models on Artificial Analysis. The model uses approximately 4.3 billion active parameters. That is 80 percent fewer than the number one open weight model. Quantized versions are coming soon for RTX Spark and DGX Spark; Jetson AGX Thor and Blackwell data center GPUs are also targets. NVFP4 versions will be deployable through NeMo-Speech.cpp or vLLM.

Liquid AI's LFM3 family provides a new generation of Liquid Foundation Models. The family spans text, vision, audio, omni and decision models built on one shared backbone. Base and post trained models are 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. These models support local workflows such as agents running in Hermes Agent, Pi and OpenClaw. They can triage inboxes and manage calendars. They can read documents and app screens. They can transcribe and translate speech. They can answer spoken questions about handwritten math. They can route support tickets or voice requests. Developers can run LFM3 models on RTX Spark with native support through llama.cpp and vLLM. They can also fine tune the models locally.

Resource management controls now appear in NVIDIA App. Users can allocate GPU resources between foreground apps and background agents. Both share the system's GPU compute. The right balance depends on what matters at that moment. Users can prioritize their foreground app for best performance. They can give their background running agent more resources to get work done faster. Access the control from System then Advanced in the NVIDIA App. This feature requires hardware accelerated GPU scheduling to be enabled in Windows Settings.

Gaming updates

Phantom Blade Zero launches tonight October 28 at 7 p.m. PT on Steam and the Epic Games Store. Players take on the role of Soul a swordsman framed for his master's murder. He travels the martial world to uncover a conspiracy. NVIDIA worked with S-GAME Studio and S-GAME Publishing to add ray traced effects. On RTX Spark laptops and GeForce RTX GPU powered PCs ray tracing adds more accurate bounced lighting. It adds detailed reflections and realistic shadows on foliage and terrain. The NVIDIA DLSS Dynamic Multi Frame Generation feature boosts frame rates with up to 6x frame generation. It can also adjust the multiplier on the fly to hit a player's target frame rate.

Version 1.6 of RTX Remix released last week. The update brought the platform to Minecraft Java Edition 1.21.1. Both creators and players can render Minecraft through RTX Remix with DLSS accelerated path tracing. They take full advantage of realistic lighting and materials in future creations. The update also adds DLSS 5 with 3D Guided Neural Rendering. It adds up to 6x Multi Frame Generation. It adds 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 support RTX Spark with DLSS 3D Guided Neural Rendering for more lifelike lighting and materials. A new Performance Mode for DLSS Frame Generation appears. New options speed up path tracing. RTX Remix 1.6 also makes AI powered modding simpler. A redesigned ComfyUI workflow lets modders generate physically based materials directly inside RTX Remix. No application programming interface setup is needed. Batch processing for full projects and presets for less technical creators are included. With NVIDIA agent skills compatible AI agents can work directly with RTX Remix to help speed up remastering projects.

What this means for different users

AI developers run models and agents locally on NVIDIA CUDA. That is the same platform that powers AI in the cloud. There is a path to deploy across more than 150 million NVIDIA RTX AI PCs and NVIDIA GPUs in the cloud. Creators 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 play hundreds of games on RTX Spark. NVIDIA DLSS ray tracing Reflex and G-SYNC deliver sharper visuals and smoother more responsive gameplay.

The Nemotron 3.5 Bolt E4B 4GB model is now available. It enables natural conversations and multistep agent tasks with as little as 4GB of memory. The model powers local voice assistants game characters and embedded interfaces on RTX Spark DGX Spark GeForce RTX GPUs and NVIDIA Jetson devices. It supports llama.cpp vLLM Ollama Unsloth and LM Studio. It works with Hermes and OpenClaw. For more demanding 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. It lets users generate and edit images locally. NVIDIA optimized versions run on RTX GPUs with as little as 8GB of VRAM. Developers can also fine tune the model on RTX Spark using an NVIDIA NeMo AutoModel recipe to customize it for specific creative workflows.

NVIDIA DGX Spark 64GB is now available to purchase from manufacturer partners. It gives developers and researchers another way to build and run AI locally. Powered by the GB10 Grace Blackwell Superchip it combines 64GB of unified memory with DGX OS and the NVIDIA AI software stack for agent development inference and fine tuning.

The announcement does not say how many of the new laptops are already in customer hands. It does not disclose injury data or safety incidents. It does not compare battery life across specific models. It does not list which of the six OEMs offer