iOS 18 boosts iPhone 15 Pro and Pro Max’s machine learning performance

Early benchmarks for the iOS 18 beta reveal significant performance gains in machine learning tasks for iPhone 15 Pro and Pro Max models. This aligns with Apple’s focus on AI advancements, as evidenced by the A17 Pro chip requirement for its new Apple Intelligence features.

With the iOS 18 beta now in circulation, early benchmarks are shedding light on the impressive performance gains, specifically on the iPhone 15 Pro models. Powered by the new software, these devices are demonstrating up to a 25% increase in Core ML Neural Engine Inference Scores compared to iOS 17.5.1.

Geekbench scores, meticulously tracked by INIYSA, reveal that while iOS 17.5.1 typically yielded Core ML Neural Engine Inference Scores in the low 6000 range for the iPhone 15 Pro Max, iOS 18 has pushed these scores to as high as 7816. This leap can be attributed in part to Apple’s introduction of the MLTensor type, optimizing compute for multidimensional arrays with a user-friendly API.

Apple further predicts notable performance boosts for specific machine learning models in iOS 18 on the iPhone 15 Pro series. These enhancements range from modest improvements like a 2% increase for MobibleNet v3 to significant leaps such as a 700% boost in performance for SSD MobileVit v2. These advancements underscore Apple’s dedication to empowering developers with cutting-edge tools and users with superior AI capabilities.

However, it’s important to note that while benchmark scores offer valuable insights into raw performance capabilities, real-world scenarios may vary. Factors such as application optimization and user usage patterns can influence actual day-to-day performance.

For users eager to explore these enhancements firsthand, the iOS 18 beta presents an opportunity to experience the latest in AI-driven advancements on the iPhone 15 Pro/Max. Early adopters are encouraged to share their experiences and feedback, contributing to ongoing discussions about the evolution of Apple’s Neural Processing Units.

Asma Hussain