A 397-Billion AI Just Ran on an iPhone
Heat trend
Collecting trend data
The percentage is based on available heat signal, not comment count or independent people.
A 397-billion-parameter AI model has successfully run on an iPhone, generating text with the network off.…
A 397-billion-parameter AI just ran on an iPhone. Not a demo video — the phone itself, generating text with the network off. On-device AI got that good, and this video shows exactly how it happened and what you can actually run today.
We break down the mixture-of-experts design that leaves 96% of the model asleep, the Apple paper that predicted all of it three years ago, why the flash-moe team deleted their own caching code and got 38% faster, and where it still falls short — speed, storage, and heat — so you know what's real and what's hype.
If you want AI that runs on your own hardware instead of somebody else's servers, this is where it's headed.
Watch Here: A 397-Billion AI Just Ran on an iPhone
#OnDeviceAI #iPhone #AI #LocalAI #MixtureOfExperts #AppleSilicon #OnDeviceAI #LocalLLM