Free Tool Now Runs Newer Google and Alibaba AI on Your Own Laptop

Most AI chatbots you use live on someone else's computers. You type, your words travel to a company's data center, and an answer comes back. llama.cpp is the opposite: it's free, open-source software that lets a regular laptop or phone run an AI model all by itself. Nothing you type leaves your machine. No monthly fee. No internet required once it's set up. It's the engine behind a lot of the "run AI privately" apps you may have seen.

The update released on September 16 is a maintenance patch, not a flashy new feature. In plain terms, the developers fixed a piece of internal math — the part that decides which words in your sentence matter most — so that newer models from Google (called Gemma) and Alibaba (called Qwen) behave correctly. Before this fix, those specific models could produce garbled or unreliable answers when running on certain setups.

The release also bundles ready-made versions for a long list of devices: Windows PCs, Apple Silicon Macs, iPhones and iPads, Android phones, Linux servers, and machines with NVIDIA, AMD and Intel graphics chips. There's even a build for Chinese server hardware. For most people, that list is the real headline — it means the same free tool works whether you're on a cheap laptop or a gaming rig.

The honest catch: this is housekeeping. If you already use llama.cpp, you will probably not notice anything different. And running AI locally still demands decent hardware — think a fairly recent laptop with 8-16GB of memory. The bigger trend is what matters: capable AI keeps getting smaller and freer, and the tools to run it yourself keep getting easier to install.