We’ve witnessed some impressive feats where devices that we’d never expected to run some form of on-device AI and successfully managed to do so, like when one person managed to operate a 3B LLM on a Pixel 7. However, it turns out smaller AI models like Gemma 3 270M can run on far less capable hardware equipped with a laughably low amount of memory. In the latest demonstration, one user ran the company’s LLM on an office printer and got decent performance out of the experiment. Using just 14.1MB, Google’s Gemma 3 270M achieved 60+ tokens per second when running […]
Read full article at https://wccftech.com/google-engineer-runs-gemma-3-270m-ai-model-directly-on-office-printer-via-postscript/
