Apple in talks with PrismML to shrink AI models for iPhone, and crypto should pay attention

Running a 27-billion-parameter AI model on a phone sounds like the kind of claim that gets you laughed out of a pitch meeting. PrismML did it anyway.

Apple is in active discussions with the Caltech spinoff to explore whether its model-compression technology can bring significantly larger AI models directly onto iPhones, according to a report from The Information. The talks center on PrismML’s ability to shrink massive language models to a fraction of their original size while preserving their reasoning capabilities.

What PrismML actually built

PrismML compressed Alibaba’s Qwen 3.6 model, a 27-billion-parameter model that normally weighs in at roughly 54 GB, down to less than 4 GB. The shrunken model ran fully functional on an iPhone 17 Pro.

The company uses proprietary mathematical techniques including 1-bit and ternary weight architectures developed at Caltech. CEO Babak Hassibi, a professor of electrical engineering at the university, leads the operation. PrismML emerged from stealth on March 31, 2026, and has already secured $16.25 million in seed funding led by Khosla Ventures, with participation from Cerberus Ventures and Caltech itself.