📱 PrismML Compresses Neural Networks for iPhone
Startup PrismML has developed a compression technology that allows Alibaba's Qwen 3.6 model with 27 billion parameters to run locally on mobile devices. The model was reduced to a size of less than 4 GB, making it possible to run on the upcoming iPhone 17 Pro. Apple has already held discussions with PrismML regarding potential integration of this technology.
🌍 Successfully localizing models of this scale (27B) on smartphones could reduce dependence on cloud computing and massive data centers, shifting the investment landscape toward Edge AI.
👤 This signifies a transition toward more private and faster AI assistants that work without the internet directly on your phone, without sending data to the cloud.