Apple is reportedly exploring new technology that could dramatically improve artificial intelligence on future iPhones. The company is interested in a startup called PrismML, whose software compresses large language models without significantly reducing their performance. If successful, the technology could allow advanced AI features to run directly on Apple’s mobile devices instead of relying on remote servers.
Apple has consistently favored on-device AI because it offers better privacy, faster responses, and reduced dependence on internet connections. However, hardware limitations have made it difficult to run the largest AI models on smartphones.
Compressing Large AI Models
PrismML has demonstrated a major technical achievement by shrinking the 54GB Qwen 3.6 language model to only 4GB. Despite the dramatic reduction in size, the company says the model continues to perform at a high level.
Large language models require huge amounts of memory because they contain billions of parameters. These parameters allow AI systems to understand complex questions, perform advanced reasoning, and complete sophisticated tasks. By comparison, smartphone AI models usually contain only a few billion parameters, limiting their capabilities.
Bringing Smarter AI to iPhones
According to reports, Apple has already met with PrismML to discuss its technology. The startup has reportedly shown that its compressed Qwen 3.6 model can run on an iPhone while handling complex conversations, advanced reasoning, and autonomous AI agents.
Apple has spent years working to reduce the size of AI models for mobile devices. However, earlier efforts often required compromises in speed or performance. PrismML’s approach appears to avoid many of those trade-offs, making it an attractive option for Apple’s long-term AI strategy.
Partnership Still Uncertain
Although discussions have reportedly taken place, neither company has announced a partnership. As a result, there is no guarantee PrismML’s technology will appear in future Apple products. Apple has also not revealed when users might see more powerful on-device AI features.
Even so, the reported talks highlight Apple’s continued focus on expanding artificial intelligence while protecting user privacy. If model compression continues to improve, future iPhones could deliver server-level AI performance without sending personal data to the cloud. Such a breakthrough would represent a significant step toward faster, more capable, and more private mobile AI experiences.
