Apple is engaged in discussions with PrismML, a Caltech spinoff, to use its advanced model-compression technology for integrating larger AI models directly onto iPhones. This effort aims to enhance on-device computing capabilities significantly.
PrismML has successfully compressed a 27-billion-parameter AI model from Alibaba, reducing its size from approximately 54 GB to under 4 GB. The compressed model has been demonstrated to operate fully functional on the iPhone 17 Pro, showcasing the potential for mobile devices to handle complex AI tasks.
Technical Innovations and Market Impact
The technology developed by PrismML employs proprietary mathematical techniques, including 1-bit and ternary weight architectures, which were pioneered at Caltech. CEO Babak Hassibi leads the company, which emerged from stealth mode in March 2026 with $16.25 million in seed funding from Khosla Ventures, Cerberus Ventures, and Caltech.
PrismML also launched Bonsai Studio, a mobile application that allows users to create images using diffusion-based techniques entirely on-device, eliminating the need for an internet connection. This capability positions the app as a unique offering in the mobile AI market.
Strategic Importance for Apple and Crypto Insights
Apple's move toward on-device AI aligns with its long-standing emphasis on privacy. By processing AI tasks locally, the company can eliminate data transmission to cloud servers, thus enhancing user security and reducing latency associated with cloud computing.
While Apple explores this technology, other tech giants are also pursuing on-device AI advancements. Google is developing its Gemini Nano models for Android, and Qualcomm has invested heavily in on-device AI through its Snapdragon processors. However, PrismML's ability to achieve a 93% reduction in model size while maintaining functionality could represent a substantial shift in the current landscape of mobile AI capabilities.
This material is informational and not financial advice.
