Apple Explores Deal With AI Startup That Shrinks Models for iPhone
Apple is in talks with PrismML, whose compression tech runs large AI models on-device using up to 15x less memory.
Apple is in active discussions with PrismML, a startup specializing in compressing large artificial intelligence models so they can run directly on an iPhone, according to a report from US Top News and Analysis. The talks signal Apple's continued push to bring more powerful on-device AI capabilities to its hardware without relying heavily on cloud infrastructure.
PrismML has developed technology capable of dramatically shrinking the footprint of AI models, with the company claiming its compressed version of Alibaba's Qwen model consumes up to 15 times less memory than the original. That kind of efficiency gain could prove critical for fitting capable AI systems within the strict power and memory constraints of a smartphone.
Read more Buffett Speeds Up Charity Giving, Targets Full Donation by 2034 →
For Apple, the potential partnership represents a meaningful step in its broader intelligence strategy. The company has faced scrutiny over how its Apple Intelligence features stack up against rivals like Google and Samsung, both of which have aggressively marketed generative AI on their own devices. Bringing a startup's compression expertise in-house — or licensing it — could accelerate Apple's ability to run more sophisticated models locally.
On-device AI processing carries significant advantages beyond raw performance: it reduces latency, lowers operating costs tied to server infrastructure, and offers stronger privacy protections since user data never leaves the device. PrismML's approach, if successfully integrated, could allow Apple to deploy larger, more capable models than its current hardware would otherwise support.
The stage of discussions and any potential deal terms remain unclear, and Apple has not publicly commented on the reported talks. Continue reading at US Top News and Analysis.
