Qwen3.8-27B reportedly runs on iPhone 18 Neural Engine at 15-70 tokens/sec
A social media post shared on Hacker News claims that the 27B-parameter Qwen3.8 model can run on the Neural Engine of Apple's iPhone 18. The reported speed ranges from 15 to 70 tokens per second, using a 1.5 bits-per-weight quantization. The source is a single linked post on X, and the claim has not been independently verified in the available material. The Hacker News thread drew 12 points and 2 comments at the time of posting.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.


Discussion (0)
Log in to join the discussion and vote.
Log in