From X (formerly Twitter) · · 30s
iPhone 18 Neural Engine Boosts AI Models
iPhone 18's Neural Engine runs Qwen 3.8-27B model at 15-60 tps.
In brief
iPhone 18's Neural Engine runs Qwen 3.8-27B model at 15-60 tps. Apple's new iPhone 18 Neural Engine processes Qwen 3.8-27B AI model at 15-60 tps. Originally reported by X (formerly Twitter).
iPhone 18 Neural Engine
Apple's new iPhone 18 Neural Engine processes Qwen 3.8-27B model at 15-60 tps.
Model Performance Details
Qwen 3.8-27B runs at 15-60 tps with base model retrained on A20 Neural Engine.
Context Processing
24K context handled with int8 K/V pairs at 70 tok/s decoding speed.
The short version
Apple's new iPhone 18 Neural Engine processes Qwen 3.8-27B AI model at 15-60 tps.
