The Free AI Engine That Runs On Your Laptop Just Got Updated

Llama.cpp is a free, open-source program that lets you run AI language models directly on your own computer or phone. Think of it as the engine under the hood: instead of asking a giant company's server to write your email, your own machine does the work. That means no monthly fee, no internet connection required, and nothing you type gets uploaded anywhere. The project is enormously popular — over 128,000 people have starred it on GitHub, the site where programmers share code.

The update released on September 13 is build b10946 — a numbered snapshot rather than a major new feature. Its headline fix is surprisingly niche: it patches how the software runs on IBM's s390x line of mainframe computers, the massive machines that still power banks, airlines and government systems. If you don't work at a bank, that specific change won't touch your life at all. It's maintenance, the software equivalent of fixing a squeaky door in a building you don't live in.

The more useful part is the packaging. Each release ships ready-made downloads for a long list of devices: Windows, Mac, Linux, Android and iPhone, plus options tuned for graphics chips from Nvidia, AMD and Intel. That breadth is what makes llama.cpp matter to non-engineers — you don't need to understand code to benefit. Free apps built on top of it let you chat with an AI offline, keep sensitive documents on your own laptop, and avoid paying $20 a month for a service that reads your messages.

The catch is real, though. Running AI on your own hardware is slower and weaker than the paid cloud versions, and the setup still isn't as simple as downloading an app from a store. Small updates like this one are also a reminder that the ecosystem moves in tiny steps, not giant leaps. But the direction is clear: private, no-cost AI is steadily becoming something a normal person can actually use.