Laya on Mac M4: Offline inference at 45 decisions per second

A user posted a demo of Laya (OS Jev) running on a Mac equipped with Apple’s M4 chip using CoreML. The model processed roughly 45 decisions per second without contacting a server. The test ran entirely offline, demons…

The demo leveraged CoreML’s quantization pipeline to fit the model into the M4’s neural engine. The author noted that the latency matched earlier GPU‑based runs, but the power draw stayed lower. No external cloud API …

ChatGPT now knows what you do on other websites via ad collector