A developer discovered their brand new iPhone 16 Pro Max couldn't perform basic math calculations using MLX LLMs, a machine learning framework. The same code ran perfectly on an older iPhone 15 Pro and a MacBook Pro. This unexpected behavior suggests a potential hardware issue within the 16 Pro Max's Neural Engine or related components.

The problem surfaced while attempting to categorize expenses using a custom app. The app, which worked flawlessly on other devices, produced gibberish results on the iPhone 16 Pro Max. Debugging revealed significant discrepancies in tensor outputs, with values off by orders of magnitude. This prompted the developer to investigate the core issue, leading to the discovery of the hardware flaw.

This incident highlights the challenges of AI on mobile devices. While Apple is pushing for on-device AI capabilities, hardware inconsistencies can lead to unpredictable results. The developer's findings raise questions about the quality control of Apple's latest devices and the reliability of their Neural Engine.

Further investigation is needed to determine the exact cause of the problem, potentially involving more in-depth hardware diagnostics. It also poses questions for developers building AI-powered apps, who must now consider device-specific quirks. The next step is likely an official response from Apple.