We learned that Apple Intelligence will require an A17 Pro chip or later at WWDC last week. And now that the iOS 18 beta is out in the wild, we’re seeing the first benchmarks for performance on the iPhone 15 Pro/Max. The new software significantly boosts Neural Engine performance with the iPhone 15 Pro Max being up to 25% faster for machine learning tasks.
Spotted by INIYSA, Geekbench scores for the Core ML Neural Engine Inference Score on an iPhone 15 Pro Max running iOS 17.5.1 have typically been in the low 6000 range.
But with iOS 18 now on devices, Geekbench scores are showing up. Notably, the iPhone 15 Pro Max running the beta software has recorded Core ML Neural Engine Inference Scores as high as 7816 – up to 25% faster than on iOS 17.
That’s an impressive jump from optimizing iOS 18 for Apple Silicon. Part of the gain in performance may be thanks to a new Core ML type that Apple calls MLTensor.
It enables compute for multidimensional arrays while using a familiar API. Here’s how Apple describes the new Core ML type:
Apple is also predicting the performance gains with iOS 18 for iPhone 15 Pro for specific machine learning models.
Those range from a small 2% improvement for MobibleNet v3, a 66% increase for ALBERT, all the way up to a 700% increase in performance for SSD MobileVit v2.
Are you running the iOS 18 beta on an iPhone 15 Pro/Max? Are you noticing an improvement in Neural Engine performance? Share your thoughts in the comments!
Catch up on everything we know about Apple Intelligence in our full guide:
FTC: We use income earning auto affiliate links. More.