Apple has unveiled Apple Intelligence 2.0, featuring a new on-device large language model that performs all AI processing locally on the iPhone 17's A20 Pro chip without sending any data to Apple's servers.
Technical Details
The 3-billion parameter model runs entirely on the Neural Engine, achieving response times under 200 milliseconds for most tasks.
- Fully offline capable: all AI features work without internet connection
- Context window processes up to 32,000 tokens (approximately 24,000 words) on-device
- Supports 12 languages at launch with more coming quarterly
- Powers Siri 2.0 with multi-step task execution and app-spanning actions
- Photo and video understanding happens entirely on-device — no cloud upload needed
Privacy Advantage
Apple is positioning the fully on-device approach as a fundamental differentiator from Google and Samsung, which rely on cloud processing for advanced AI features. Apple's VP of machine learning stated that private AI is not just a feature but a right, drawing a clear competitive line in the AI smartphone race.
