Free AI You Run Yourself Just Got a Tune-Up for Phones and Laptops

llama.cpp is free software that lets you run AI chatbots on your own computer or phone instead of paying a company like OpenAI or Google to do it in the cloud. Think of it as the engine, not the car — other apps and tools use it to power their private, offline AI. This week the project shipped a small update with the unglamorous name "b11306." Its main change is a fix to an internal test that checks whether AI models are behaving correctly. In plain terms: it's a quality-control tweak, not a new feature.

So why mention it at all? Because the release notes read like a list of everything the software now supports. There are versions for Apple Silicon Macs, Intel Macs and iPhones; for Windows machines with NVIDIA or AMD graphics cards; for Linux servers; and — notably — for Android phones running Snapdragon chips, where it can use the phone's graphics processor and its dedicated AI chip. The project also publishes an app and setup guides, so getting started is easier than it used to be.

That spread matters because running AI on your own device means your messages, documents and photos never leave your hands. No company sees your prompts, no monthly subscription, and it keeps working when your Wi-Fi doesn't. The trade-off is that smaller on-device models are generally less capable than the biggest cloud ones, and battery and storage take a hit.

The catch is the audience. Releases like this are aimed at developers and hobbyists, not casual users. There's no app-store download-and-go experience; you may need to follow a setup guide, pick a model file (often several gigabytes) and accept that things can break. If you're curious, search for "llama.cpp" plus your device name — but expect a weekend project, not a five-minute install.