Google AI Edge Gallery Brings Gemma 4 to Phones: Run LLMs Fully Offline with 256K Context
Google has officially released AI Edge Gallery on both Google Play and the Apple App Store, enabling anyone to run the Gemma 4 model entirely on-device without an internet connection. This marks a significant step toward mainstream on-device AI.
What Is AI Edge Gallery?
AI Edge Gallery is a local sandbox for Google's Gemini models. Unlike cloud-based Gemini, it downloads the AI model directly to your device -- meaning full offline capability and enhanced privacy.
Gemma 4 Highlights
- Architecture: Built on the same foundation as Gemini 3
- Context window: 256K tokens
- Improvements: Better logic, multilingual support, and fluid conversation
- Requirements: Android 12+ or iOS 17+
Key Features
- Prompt Lab: Summarize large PDFs or write complex code locally
- Ask Image: Identify objects, plants, or text in photos without server connection
- Audio Scribe: Offline transcription
- Fluid conversation: No interruptions from poor internet connections
Model Tiers
- Gemma E4B: For flagship devices (Pixel 10 Pro XL) -- best for long documents, coding, complex planning
- Gemma E2B: For mid-range phones or faster experience
Why It Matters
Until now, running flagship AI models on-device required sideloading APKs and technical setup. By moving to mainstream app stores, Google is signaling that on-device AI is ready for prime time.
The privacy implications are significant: all processing stays on your device. No data sent to Google servers. This is particularly valuable for sensitive tasks like document summarization, medical information processing, and private conversations.
On-device AI also enables use cases that were previously impossible: AI assistance on flights, in remote areas, or in environments where cloud connectivity is restricted or prohibited.