The Free Tool That Runs AI on Your Laptop Just Got an Update

If you've ever paid a monthly fee to chat with an AI, there's a free alternative you may not know about. It's called llama.cpp, and it lets you download an AI model once and run it right on your own laptop or desktop — no subscription, no internet connection, no company watching your questions. This week the project pushed out a new build, numbered b11081, which is essentially a maintenance release rather than a splashy new feature.

So what actually changed? Mostly plumbing. The developers rewrote parts of their automated testing setup — the behind-the-scenes robots that check whether new code breaks anything — and added extra command-line options, including one that lets you filter which AI model types get tested. They also switched some checks to a more precise error measurement and cleaned up log messages. Boring on purpose: this is the unglamorous work that keeps software from randomly crashing.

The bigger story is what llama.cpp represents. Because it's free and open source — meaning anyone can inspect and improve the code — it has become the engine inside dozens of homegrown AI apps. It runs on Macs, Windows PCs, Linux servers and even Android phones. That's why the release notes list builds for Apple Silicon, Nvidia and AMD graphics cards, and chips from Intel, Qualcomm and Huawei. Someone with a mid-range laptop from the last few years can often run a capable chatbot locally.

The catch: local AI is still slower and less capable than the biggest cloud models, and setup isn't exactly one-click. You'll need to pick a model, download several gigabytes, and follow instructions. But for private notes, drafting, or offline use on a plane, it works — and it costs nothing per month. This release makes that experience a little less fragile.