I Ran a Local LLM on My Phone for a Week. It’s Closer Than You Think.
中文摘要
iPhone 15 Pro运行本地30亿参数模型可达每秒30 token，显示手机本地AI大模型已趋于实用。
English Summary
Running a 3B parameter LLM on iPhone 15 Pro reaches 30 tokens per second, showing that viable on-device AI is closer than expected.
原文节选
Thirty tokens per second. That’s the generation speed Apple’s on-device foundation model hits on an iPhone 15 Pro: a ~3 billion parameter… Continue reading on Medium »