NVIDIA RTX Spark laptops launch Oct. 16 with local AI agent tools and apps

Local agents set up in a few clicks

NVIDIA describes RTX Spark as a single superchip that combines the capabilities of a gaming PC, a creative studio and a personal AI supercomputer. NVIDIA and Microsoft announced last week that RTX Spark laptops will be available on Friday, Oct. 16.

From the first boot, users will see a "Getting Started" screen where they can choose and install an agent app. According to NVIDIA, agents will run in a sandbox built on NVIDIA OpenShell and Microsoft Execution Containers (MXC). MXC provides operating system-level security, while OpenShell makes it simpler for agents to adopt MXC and adds credential protection and security monitoring. OpenClaw and GitHub Copilot already support MXC, and Hermes Agent will integrate OpenShell controls and MXC by the end of the month.

GitHub Copilot, Hermes Agent, OpenClaw, Perplexity, Pi and WorkBuddy will recommend an optimized local model and settings for RTX Spark and configure them in a few clicks. For users who want both local and cloud models, NVIDIA Switchyard, an open source model router library, lets agents route queries between the two. Hermes will add it later in October, while Perplexity Computer and GitHub Copilot have built their own routing.

NVIDIA also says it has released MCPs for its applications, including NVIDIA App, Broadcast, PAIR, Omniverse and RTX Remix. AI playbooks for RTX Spark arrive on Friday, Oct. 16, covering tasks such as setting up a first agent or building an AI chief of staff.

Nemotron 3.5 Bolt 74B and PAIR

NVIDIA Nemotron 3.5 Bolt 74B-A9B 24GB arrives on Thursday, Oct. 15. NVIDIA calls it an LLM purpose-built for local agents on RTX Spark. It is a mixture-of-experts model with 74B parameters and 9B active parameters, packed into 24GB using 2-bit quantization and quantization-aware training. NVIDIA says it is faster than most popular dense models and smarter than most popular MoE models with a similar memory footprint.

At launch it can run through llama.cpp, vLLM, Ollama, Unsloth and LM Studio, and agents such as Hermes and OpenClaw can access it directly. NVIDIA says the model is fully open source, including training data, weights and recipes.

NVIDIA PAIR, a personal AI router, moves out of beta. The free, open source tool discovers compatible PCs on a local network, installs models and inference engines, and distributes AI workloads across them. The release adds llama.cpp support, direct Hugging Face downloads, customizable inference engine settings and RTX Spark support. It works on Windows, Linux and macOS.

Creative apps, developer tools and gaming

NVIDIA says Adobe is optimizing Photoshop, Premiere and Substance 3D for RTX Spark, with up to 2x faster AI, editing, color and effects. Photoshop gets GPU-accelerated compositing, and Premiere speeds up playback, rendering, color grading and features such as Enhance Speech and Object Mask. AI optimizations in Photoshop and Premiere are in beta now, and Substance 3D updates are coming soon.

NVIDIA Broadcast 2.3 is available now and runs natively on RTX Spark. It adds RTX Video Super Resolution and Frame Generation effects, Studio Voice profiles, an Offline File Processor for prerecorded content, and an MCP server so local agents can adjust effects and process files. RTX Video Frame Generation and HDR are available now in ComfyUI, and DLSS Ray Reconstruction comes to Blender 5.3 in November.

For developers, PyTorch, Python, CUDA and all CUDA libraries are available on RTX Spark, and GitHub Copilot, Pi and other coding agents are optimized for it. Llama.cpp adds support for decision models and releases a Windows app.

For games, NVIDIA says hundreds of titles and anti-cheat are supported. A 2x faster DLSS 5 AI model arrives Oct. 16, and a Performance Mode for DLSS Frame Generation is available via the NVIDIA App beta. Nemotron 3.5 Bolt E4B 4GB, a small language model for NVIDIA ACE game characters, and an RTX Remix update with DLSS 5 support both arrive Friday, Oct. 23.

What it means for users

NVIDIA argues that local models have become capable enough to power agents, noting that Qwen 3.8-Flash-Next is catching up to frontier cloud models released a few months ago and that local LLM downloads are already 5x higher than all of last year. The company's launch package is aimed at making that practical on a laptop: guided agent setup, sandboxed security, hybrid routing and an open model sized for the hardware. Buyers of RTX Spark laptops should find these tools available at or soon after launch, with some items, such as the Adobe Substance 3D updates, still to come.