OpenAI Whisper has 12 unique checkpoint files, but the Python package exposes 14 local model names due to aliases for 'large' and 'turbo'. Additional confusion arises from runtimes and converted forma…
Researchers reverse-engineered the Apple Neural Engine (ANE), documenting its architecture, programming interface, and performance across A11-A18 and M1-M5 chips. The study reveals a direct user-space…
Apple is skipping the high-end M6 Pro and Max chips to fast-track its M7 family, aiming to deliver a massive leap in memory bandwidth and Neural Engine capacity for on-device AI. The base M7 will offe…
Duration AI successfully ran PrismML's Bonsai Image model on a 2020 iPhone 12 Pro using mlx-swift, generating 512x512 images in two to three minutes despite thermal throttling. The prototype demonstra…
Apple introduced Core AI at WWDC 26 as the successor to Core ML, designed to run large language models and generative AI on-device using Apple Silicon. The framework segments Apple's ML ecosystem into…
Apple announced Core AI at WWDC26 as the successor to Core ML, a developer framework for running large language models and generative workloads optimized for Apple Silicon with on-device performance a…
Apple announced Core AI, a new framework for running generative AI models on-device, at WWDC 26. The framework supports large language models up to 70B parameters and runs exclusively on Apple Silicon…
A developer built MailBroom, an iPhone email cleaner that runs AI entirely on-device, claiming it is 88% greener than cloud alternatives. The app processes emails locally using Core ML, avoiding data …
Apple's on-device AI strategy represents a privacy-first, performance-oriented architectural break from cloud-centric AI, enabling local-first applications where user data never leaves the device and …
The W3C draft specification WebNN defines a graph-based neural network API that enables JavaScript to submit static model graphs to native OS runtimes, according to a November 21, 2025 blog post on zi…