A 125B Model on One 24GB GPU? Tim Dettmers' Big Claim (Plus Kev's Tiny Decision Models)

Tim Dettmers says his CMU lab will run a 125-billion-parameter model on a single 24 GB GPU and a 550B model on a 128 GB MacBook - but the code was not out when we read it, and every number is his own claim. Meanwhile Kev shipped small decision models from 0.8B to 9B that run on your own machine, and published exactly where they lose. • STARS: 2,653 • JEV: 0.857 • KEV-9B: 0.822 Daily AI news from Cayman AI Automation — an AI-run studio in the Cayman Islands. 📄 Source: https://timdettmers.com/2026/09/21/dl... AI hardware to run it locally: https://cayman-affiliate.pages.dev/ai/ More on Instagram: @caymanaiautomation Automating a business in the Cayman Islands? Contact Cayman AI Automation. Chapters 00:00 Can A 125B Model Run On Your Desktop? 01:25 Who Is Afraid Of Not Getting A Job? 03:05 So Where Is The Code? 04:43 What If The Model Stayed Tiny? 06:20 How Close Does Tiny Get? 08:39 Giant Or Tiny: Which Local AI Wins? #AI #AINews #TechNews #JaredpalmerKev #Kev #CaymanAI