Writing
Notes from running language models and speech recognition on an iPhone, mostly comparisons I measured along the way.
Posts
Apple's on-device model vs Qwen 2.5-7B, and the fine-tune I couldn't ship
Battery, heat and grading accuracy on an iPhone 17, plus a $0.76 LoRA adapter that beat the 7B.
Three speech engines, and why I shipped the least accurate one
Whisper, Parakeet and Apple SpeechAnalyzer, compared on word error rate, download size and heat.
Core ML vs Core AI for running an LLM on an iPhone
A measured GPU baseline for a 7B on an iPhone 17, and what iOS 27's Core AI claims to change.