The Free AI That Runs on Your Own Laptop Just Got Fixed

Llama.cpp is a free, open-source program that lets you run AI chatbots directly on your own laptop, desktop, or phone — no monthly subscription, no internet connection, and no company reading your conversations. Think of it as the engine under the hood of many private AI apps. A new version, labeled b11138, was released on September 23rd, and it's a small but worthwhile tune-up rather than a flashy new feature.

The fix deals with something called Jinja templates. In plain English, that's the set of formatting rules AI models use to decide where your question ends and the AI's answer begins. Some of those rules use plus and minus signs — for example, telling the AI to count backward a certain number of items from the end of a list. The old code got confused about when to apply those signs, which could break certain models or produce garbled results.

This update teaches the software to handle those plus and minus signs in the right order. If that sounds tiny, that's because it is — but tiny fixes like this are exactly what keep free, private AI usable. The release also includes ready-made versions for a wide range of hardware: Macs, Windows PCs, Android phones, and even specialized chips from Nvidia and Qualcomm. That matters because it means you don't need an expensive new computer to try AI that stays on your device.

So what's the takeaway? Big AI news usually comes from companies like OpenAI or Google. But the quieter story is that free, do-it-yourself AI is maturing, one small patch at a time. If you've ever worried about pasting sensitive information into a chatbot, tools built on llama.cpp are the alternative — and they just got a little more dependable.