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

Llama.cpp is a free, open-source piece of software with a cult following — nearly 129,000 people have starred it on GitHub, the site where programmers share code. Its whole purpose is simple and a little radical: let ordinary computers run AI chatbots locally, without sending your questions to a company's servers. That means no monthly bill and no one else reading your conversations.

This release, labeled b11060, is a maintenance update rather than a headline feature. It fixes a small technical bug affecting a family of AI models called Mamba. In plain terms, computers work fastest when data sits in one tidy, unbroken block in memory — think of a librarian pulling a single book off a shelf versus hunting through twenty scattered piles. The fix makes sure one part of the Mamba model keeps its data in that tidy block, and it removes an unnecessary copying step afterward. Boring? Yes. But small efficiencies like this add up, especially on modest hardware like a phone or a cheap laptop.

Interestingly, the notes credit ChatGPT as an assistant on the change — a reminder that AI is increasingly used to help write and improve other AI software. The release also ships ready-made versions for Macs, Windows PCs, Linux machines, Android phones and iPhones, including builds that tap into graphics chips from Nvidia and AMD for extra speed.

So what does this mean for you? Mostly nothing today, and that's fine. It's the unseen plumbing work that keeps a free alternative to paid AI subscriptions healthy. If you've ever wanted to try an AI assistant without a credit card or a privacy worry, this project is where that starts — and it keeps getting a little smoother.