Microsoft announced Project Zenith on Sept. 4: a preconfigured Windows 11 setup for developer machines with at least 64 GB of unified memory, according to the Windows Developer Blog.
Machines need at least 250 GB per second of memory bandwidth to qualify, Microsoft said, a bar meant to support running local AI models with more than 30 billion parameters without metered cloud usage.
The setup pins Windows Terminal and Visual Studio Code to the taskbar, integrates Windows Subsystem for Linux with container support, and installs tools "spanning languages and runtimes, source control, and productivity tools," Microsoft said. File Explorer switches on hidden files, file extensions and full paths by default, while Start menu tips, account notifications and sync prompts are turned off.
Project Zenith launches first on devices built around AMD's Ryzen AI Halo chips, with more hardware from other partners coming "in the coming months," Microsoft said.
Running a 30-billion-parameter model locally still means owning hardware with server-class memory bandwidth. Microsoft is packaging that requirement as a product category here, not lowering the bar to reach it.