Tag
A user reflects on why more apps don’t run local LLMs directly on phones, noting Gemma 2-4B models already work offline and could eliminate server costs while maintaining near-GPT-4o quality.