Apple is exploring a startup called PrismML, which works on AI compression technology for mobile devices. The startup says it can cut huge AI models down by using up to 15 times less memory, helping them run on iPhones instead of remote servers. According to reports, PrismML has already compressed Alibaba’s open-source Qwen model from about 54 GB to under 4 GB. The company also says the smaller model can run all 27 billion parameters locally on an iPhone 15 or newer.
That matters because on-device AI can improve speed and privacy while reducing energy use. PrismML claims its compressed models generate answers six to eight times faster and consume three to six times less energy than traditional versions.
The talks are still early, and Apple has not announced any deal. PrismML CEO Babak Hassibi told CNBC that Apple and other companies are evaluating the technology right now.
For Apple, this could be a useful step in its AI plans. If the technology works well inside iPhones, users could get more powerful AI features without sending as much data to the cloud.
FAQs [Frequently Asked Questions]
1. What is PrismML?
PrismML is a startup that compresses large AI models so they can run on phones, especially iPhones, with less memory, faster speed, and lower energy use.
2. Why is Apple interested in it?
Apple wants stronger on-device AI, which can improve privacy, reduce cloud dependence, and make AI features work faster on iPhones.
3. Has Apple made a deal yet?
No deal has been announced. The reports say Apple is only in early talks and is still evaluating PrismML’s technology.