Edge inference will start becoming a thing next year. Apple is clearly working towards this. I asked some LLMs when they think we’ll be able to run a GLM-5.2 grade model locally on a smartphone. Here’s what they said: GLM-5.2: Nov 2027 GPT-5.6 Sol: Sep 2028 Grok 4.5: Oct 2028 Fable 5: Mar 2029 Curious to see how this impacts inference providers 2028 and beyond.(Alex Svanevik 🐧)