Yhn TechnologyMobile first seen 3 d ago, last 1 h ago, peak #8
Qwen 27B model runs on iPhone Neural Engine at up to 70 tokens per second
Original: Qwen3.8-27B on the iPhone 18 Neural Engine at 15–70 tok/s (1.5 bpw)
A 27-billion-parameter Qwen language model has reportedly been run on an iPhone 18's Neural Engine, achieving generation speeds of 15 to 70 tokens per second using extreme 1.5-bit weight quantization. The result suggests current flagship phones can host large open models entirely on-device, a development drawing attention from the local AI and edge-inference communities.
Why now: On-device inference of a large language model on consumer phone hardware is a notable technical milestone for local AI.
Evidence
API: https://socialmediatrends-api.osmike.com/v1/trends/2642601