The Free Tool That Runs AI on Your Own Laptop Just Got a Fix

Llama.cpp is a free, downloadable program that lets you run AI language models directly on your own laptop or desktop — no subscription, no internet connection required, and no company watching what you type. Think of it as the difference between cooking at home and ordering takeout every night. The project, started by developer Georgi Gerganov, has been starred on GitHub over 130,000 times, making it one of the most-followed AI projects in the world.

The update released this week, labelled b11342, is not a flashy new feature. It fixes a bug where the software could fail to create its cache folder — the place it stores temporary files — when your computer's folder structure uses shortcuts called "symlinks." That failure happened on certain older versions of a common Linux system library. In practice, some users on Linux, and anyone using the tool's remote-connection feature, may have seen the program stumble or refuse to start.

The fix was contributed by Adrien Gallouët, a developer at Hugging Face, the popular AI model-sharing site. The project also shipped ready-made versions for a huge range of hardware: Macs with Apple Silicon, Windows PCs with NVIDIA or AMD graphics cards, Android phones, and even Snapdragon chips with dedicated AI processors. That matters because it means ordinary people can squeeze AI performance out of the devices they already own.

Why should a non-technical person care? Because llama.cpp is quietly the backbone of the "run AI at home" movement. If you've used a private AI assistant app, a local coding helper, or a chatbot that works on a plane with no Wi-Fi, there's a decent chance this project is under the hood. Small fixes like this one are what keep that alternative alive and dependable.