Open LLM weights, running on your phone

Open model weights changed the game: the same breed of AI models behind cloud chatbots now runs directly on an Android phone - here is how.

What open weights actually means

A cloud chatbot rents you answers; an open-weight model hands you the model file itself. Llama popularized the idea, and families like Qwen, Gemma, Phi and LFM followed - small models you can download and run anywhere.

What your phone needs

Three things: an app that runs models, a model file sized for your device, and enough free RAM. Models from 0.5B parameters run on modest phones; 12B models want flagship hardware and plenty of storage.

The easy route

Local AI does the plumbing for you: a model browser with one-tap downloads, then streaming chat with whatever you installed. After the download, no internet, no account, no server in the loop.

Setting expectations

A phone-sized model is brilliant at drafts, summaries, brainstorming and conversation - not at everything a data center can do. Start with a small model, then step up until quality and speed feel right.