iPhone 17 Pro Demonstrated Running a 400B LLM

The iPhone 17 Pro runs a 400-billion-parameter LLM locally, a feat earning 328 points and 189 comments. With on-device inference achieving 45 tokens per second, it beats the average edge device by 3x. The demonstration highlights that mobile hardware now exceeds datacenter capabilities from just three years ago, making privacy-preserving AI a tangible reality for consumers.