MiniMax
On-Device Inference: From 596 Seconds to 10, but Not Through Smaller Models
An on-device screenshot-recognition optimization cut latency from 596 seconds to 8–11 seconds through runtime, GPU, and output changes.
Back to home
on-device models
2 articles tagged with this topic
Google
Chrome Silently Installs 4GB AI Model: Google Races Ahead in Local AI via Browser
Chrome silently installs a ~4GB local AI model without consent. Browsers are becoming AI runtimes—distribution rights now matter more than the models.