Apple is in talks with startup PrismML about running larger AI models directly on iPhones, potentially allowing 27 billion active parameters compared with the 1 billion to 4 billion currently active in Apple's on-device model. Larger on-device models could reduce Apple's reliance on cloud servers and improve user privacy by keeping more AI processing local to the device.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Apple has held meetings with startup PrismML about technology to run larger AI models on iPhones. PrismML has shrunk Alibaba's Qwen 3.6 model—which has 27 billion parameters—to run entirely on an iPhone 17 Pro, compared with Apple's current on-device AFM 3 Core Advanced model which has 20 billion parameters.
Why it matters
Unlike Apple's current model, which uses a sparse architecture where only 1 billion to 4 billion parameters are active at a time, all 27 billion parameters in PrismML's model can be active simultaneously. Running larger, fully active models on-device could reduce Apple's server costs and enhance user privacy by shifting more Apple Intelligence features away from Apple's Private Cloud Compute servers.
What to watch
The talks remain exploratory at this stage. If Apple adopts PrismML's approach, it would represent a significant shift in how the company delivers AI features to iPhone users, potentially enabling more sophisticated on-device AI capabilities.
Apple's current approach to on-device AI relies on a sparse architecture for its AFM 3 Core Advanced model, which powers iOS 27 features like Siri's more expressive voices and improved dictation on iPhone 17 Pro and iPhone Air models. The sparse design keeps computational load manageable by activating only a fraction of the model's parameters at any given time. PrismML's breakthrough—successfully compressing a 27-billion-parameter model to run fully on an iPhone—suggests an alternative path: deploying larger models with all parameters active, potentially delivering richer AI capabilities without leaving the device.
The significance for Apple lies in both economics and privacy. Moving more AI processing from Private Cloud Compute servers to the device would reduce infrastructure costs and eliminate the need to transmit user data to Apple's servers, directly addressing privacy concerns. The talks remain preliminary, but they signal Apple's interest in expanding the sophistication of on-device intelligence while maintaining the privacy guarantees its users expect.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
