Windows is being optimized for Nvidia RTX Spark laptops, with local AI and improved app compatibility

Windows adds local AI agents

RTX Spark laptops are built around an Arm CPU, a Blackwell GPU and unified memory, a combination intended to run AI models on the device. Microsoft has worked to make starting with AI agents straightforward.

A Windows “Get Started” flow helps users find and install an agent, according to Microsoft program manager Peter Dawoud. He said users can reach an agent in two clicks. Agents can also be installed and updated through the Windows Store.

Hermes, Perplexity and OpenClaw are the first supported agents at launch. Microsoft is working to add more; Meta’s Muse is expected to come to Windows in the future.

Setup and speed vary

Testing found that Hermes was mostly a two-click setup. OpenClaw, however, reported that a local gateway was not installed correctly and needed to be removed and installed again.

Local AI models may be useful for coding and background tasks, but they can take longer to answer prompts than cloud models. More testing is needed to assess their value beyond straightforward tasks.

Prism gets platform tuning

Apps without native Arm versions rely on Microsoft’s Prism emulator on RTX Spark laptops. Windows on Arm struggled with emulation for years, but Arm-native apps and improvements to Prism have helped the new systems.

Dawoud said emulation requires adjustments for each platform and that Microsoft optimized x64 translation for the RTX Spark chip’s microarchitecture. The work also targets gaming, where CPU performance can affect frame pacing and how smooth a game feels. Microsoft ships Prism updates monthly to improve compatibility, Dawoud said.