Microsoft lets users delete its 2.5GB Phi Silica AI from Windows 11
Microsoft finally realized that not everyone wants their hard drive stuffed with heavy AI models they never asked for. The latest experimental update brings a magical button that solves a massive headache for the new wave of "smart" computers.
The hidden option popped up in the experimental preview build 26300.8553 of Windows 11 for Insiders. Tech enthusiasts digging through the settings menu discovered a new, glorious "Uninstall" button. This button targets Phi Silica, a local small language model that Microsoft silently bundled into modern operating systems.
To run these fancy local AI tricks, the system requirements for Copilot+ PCs had to be aggressively bloated. While a basic machine needs just 4 GB of RAM and 64 GB of storage, these AI-powered rigs demand a hefty 16 GB of RAM and 256 GB of disk space to accommodate a neural processing unit.
A massive chunk of this extra hardware tax goes directly to housing local models. The Phi Silica component alone eats up over 2.5 GB of disk space just sitting there, waiting to help rewrite or summarize text.
To make things even busier, Microsoft also introduced another built-in micro-model called Mu, which helps navigate system settings using natural language. This helper was trained on data generated by Phi Silica, making the operating system look more like a crowded AI dormitory than a clean workspace. The new uninstall toggle allows users to completely wipe the entire model with a single click.
Selling overpriced computers with inflated system requirements just to force-feed users local AI models that they end up immediately deleting is peak tech industry logic. It seems the most anticipated AI feature of the year is simply the ability to turn it off.
Source: Neowin
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.