Free App Lets You Run AI Chatbots On Your Own Laptop
No monthly fees, no internet needed — and your private data never leaves home.
Most people use AI through a website or app, which means their questions and documents travel to a company's computers somewhere else. llama.cpp is free, open-source software that flips that around. It lets you download an AI model once and then run it directly on your own laptop or desktop — like keeping a cookbook at home instead of calling a restaurant every time you want dinner. The project just published build b10989, a routine but useful update.
The main technical change is a better way for the software to share work across multiple AMD graphics cards, a feature called AllReduce for ROCm. Translated: if you own a PC with AMD hardware, the AI can now split a big job across several of its chips and finish faster. ROCm is AMD's answer to Nvidia's CUDA, the toolkit that lets graphics cards do AI math instead of just drawing video games.
Just as important is what came along for the ride: ready-to-run downloads for a long list of systems. There are versions for Windows (both regular and ARM laptops), Macs with Apple Silicon, Linux machines, and Android phones. Some come pre-built for specific graphics cards. That matters because the number one reason people give up on local AI is that installing it is a headache. Pre-built packages remove most of that pain — you download, you double-click, you're chatting.
So what's the catch? Local AI is not magic. The bigger, smarter models need a decent computer with a fair amount of memory, and a mid-range laptop will feel slow compared with ChatGPT. You also won't get the newest, largest models, which stay behind company servers. But for everyday tasks — summarizing documents, drafting emails, translating — a small model running privately on your own machine is often good enough, and it costs nothing after setup.
- llama.cpp is free software that runs AI chatbots directly on your computer, so nothing you type gets sent to a company's servers
- This update improves support for AMD graphics cards, which can mean faster responses for people with those PCs
- New ready-made downloads for Windows, Mac, Linux, and Android make setup much easier than before — though you still need a reasonably capable machine
Why It Matters
You can run capable AI privately and free, without subscriptions or sending personal data to big companies.