on device

Asked on 2025-06-30

1 search

Apple's WWDC sessions have highlighted several advancements in on-device capabilities. Here are some key points:

  1. Foundation Models Framework: Apple introduced a large language model with 3 billion parameters, optimized for on-device use. This model supports tasks like summarization, extraction, and classification, ensuring user data privacy as it operates without needing server-side processing. This is part of the broader Foundation Models framework, which integrates with Apple's suite of machine learning APIs for on-device intelligence. For more details, you can refer to the session Meet the Foundation Models framework.

  2. SpeechAnalyzer: The new on-device model for speech-to-text, called Speech Transcriber, allows developers to integrate advanced speech recognition capabilities into their apps without increasing app size or memory usage. This model is designed to keep speech data private and is available across all platforms except watchOS. More information can be found in the session Bring advanced speech-to-text to your app with SpeechAnalyzer.

  3. Power Profiling: The Power Profiler tool is available on-device, allowing developers to diagnose battery usage issues that may not be reproducible during development. This tool helps in gathering real-world data without needing a direct connection to Xcode. For more insights, see the session Profile and optimize power usage in your app.

These sessions illustrate Apple's focus on enhancing on-device processing capabilities, ensuring privacy, and providing developers with robust tools to optimize app performance.