Free Tool Lets You Run Powerful AI on Your Own Laptop

Most AI chatbots you use — ChatGPT, Gemini, Claude — live on huge servers owned by big companies. You pay a monthly fee or hand over your data, and every question travels across the internet. A free project called llama.cpp takes the opposite approach. It's open-source software that lets an ordinary computer run AI models by itself, with nothing sent anywhere. This week the team behind it pushed out build b11042, a small but genuinely useful update.

The headline change is about graphics chips. Modern laptops and phones have small processors built for drawing images, and it turns out they're also good at the maths AI needs. This update improves llama.cpp's support for OpenCL (a way for software to borrow your graphics chip for non-graphics work). Specifically, it adds a faster method for one particular AI calculation on Qualcomm's Adreno graphics — the kind found in many Windows laptops and Android phones. In plain terms, it helps those devices do more of the work in fewer steps.

Why should you care? Because the gap between "AI runs in a distant data centre" and "AI runs on the thing in my bag" is closing. Local AI means no monthly bill, no internet required, and your private notes, medical questions or work documents never leave your device. That's a real difference for anyone uneasy about handing personal information to a tech giant.

The catch: this is an engineering patch, not a headline breakthrough. Most people will notice nothing at all. Running AI locally still needs a reasonably modern machine, some fiddly setup, and patience while it thinks. But each small release like this one widens the pool of devices that can do it — and that's the trend worth watching.