AI Edge Gallery is the premier destination for running the world’s most powerful open-source Large Language Models (LLMs) on your mobile device. Experience high-performance Generative AI directly on your hardware—fully offline, private, and lightning-fast.
Now Featuring: Gemma 4
This update brings official support for the newly released Gemma 4 family. As the centerpiece of this release, Gemma 4 allows you to test the cutting edge of on-device AI. Experience advanced reasoning, logic, and creative capabilities without ever sending your data to a server.
Core Features
- Agent Skills: Transform your LLM from a conversationalist into a proactive assistant. Use the Agent Skills tile to augment model capabilities with tools like Wikipedia for fact-grounding, interactive maps, and rich visual summary cards. You can even load modular skills from a URL or browse community contributions on GitHub Discussions.
- AI Chat with Thinking Mode: Engage in fluid, multi-turn conversations and toggle the new Thinking Mode to peek "under the hood." This feature allows you to see the model’s step-by-step reasoning process, which is perfect for understanding complex problem-solving. Note: Thinking Mode currently works with supported models, starting with the Gemma 4 family.
- Ask Image: Use multimodal power to identify objects, solve visual puzzles, or get detailed descriptions using your device’s camera or photo gallery.
- Audio Scribe: Transcribe and translate voice recordings into text in real-time using high-efficiency on-device language models.
- Prompt Lab: A dedicated workspace to test different prompts and single-turn use cases with granular control over model parameters like temperature and top-k.
- Mobile Actions: Unlock offline device controls and automated tasks powered entirely by a finetune of FuntionGemma 270m.
- Tiny Garden: A fun, experimental mini-game that uses natural language to plant and harvest a virtual garden using a finetune of FunctionGemma 270m.
- Model Management & Benchmark: Gallery is a flexible sandbox for a wide variety of open-source models. Easily download models from the list or load your own custom models. Manage your model library effortlessly and run benchmark tests to understand exactly how each model performs on your specific hardware.
- 100% On-Device Privacy: All model inferences happen directly on your device hardware. No internet is required, ensuring total privacy for your prompts, images, and sensitive data.
Built for the Community
AI Edge Gallery is an open-source project designed for the developer community and AI enthusiasts alike. Explore our example features, contribute your own skills, and help shape the future of the on-device agent ecosystem.
Check out the source code on GitHub:
https://github.com/google-ai-edge/gallery
Note: This app is in active development. Performance is dependent on your device's hardware (CPU/GPU). For support or feedback, contact us at google-ai-edge-gallery-android-feedback@google.com.
[TL;DR] Great app, but some bugs need to be addressed. Could use .safetensors model support as well.I honestly think this app is genius. Let’s say that you’re out in the middle of nowhere and you need to see what type of food is safe to eat. You upload an image, and let the Gemma model do its thing. Or, when you’re back online, you can use the Agentic mode where it can access the internet (with custom plugins).You might be wondering why this review it only 4 stars. The bad part: the app is super buggy on an iPhone 15. After a second prompt, the app freezes. Want to change the model’s settings? Nope; another crash. And the problem with this app is that you can only use their special litertlm format for the models (the reason why is that it manages the orchestration, kv cache, and memory mapping all in one). It would be great to see safetensors support.Also the desktop app is bugged as well. It’s just the iPhone app but ported to Mac. So it’s just an iPhone simulator (basically). Just don’t use this on desktop; use ollama, lm studio, or vllm instead. In conclusion, this app has great potential, but would need some improvements.
Why should I be complaining
Neb Mubabbelppap
I am using this app on an iPhone 14. I thought this was like Gemini and upon downloading and starting I was pretty disappointed. The AI didn’t seem that good and the app kept freezing after a couple prompts. When I looked into it, I realized this was actually Google‘s Gemma four model instead of running in a data center around on my device, which was mind blowing. Even with the freezing. I’m gonna take this experience like a plane. You can’t really complain when you’re flying 4000 feet in the air 12 times faster than anything on the ground. In AI model that runs on my device off-line is exactly like that. I’m excited to see where this type of projects by Google go!
It’s local, and that’s the best part
Alex_2027_
Yes, it is smart. If you’re on a smaller iPhone with less RAM, it knows which models are available, and the models do have a lot of baked intelligence. I tried to use the one for iOS specific actions, there’s still barely any functionality. The pre-created prompts that are listed are almost all that it can do. I tried opening some apps, I tried creating an email. I tried changing any settings. It was basically utterly useless.Small local prompts that don’t memory are basically the best they can do at this point. And the worst part is there is no widget available to quickly call while still on the home screen. You actually have to go into the app in order to do a basic function that is faster to do with the widgets. It’s still a solution looking for a problem.
Nice app, but privacy concerns
EnchantedTortilla
It works pretty well, and lower memory requirements, depending on the task, than loading the full model in other frameworks But: Why does a local AI app which runs on device (purportedly) give me the functionality to thumbs up or thumbs down the response??! personally, if I’m running it on device and powering the matmuls by draining my battery I want the model running in a Faraday cage
- Added global translation and localization support across 100 different countries.
- Bug fixes.
Version 1.0.10
The developer, Google, indicated that the app’s privacy practices may include handling of data as described below. For more information, see the developer’s privacy policy .
Data Linked to You
The following data may be collected and linked to your identity:
Identifiers
Diagnostics
Other Data
Data Not Linked to You
The following data may be collected but it is not linked to your identity:
Location
Usage Data
Diagnostics
Privacy practices may vary, for example, based on the features you use or your age. Learn More
Accessibility
The developer has not yet indicated which accessibility features this app supports. Learn More
Information
Seller
Google LLC
Size
98.4 MB
Category
Productivity
Compatibility
Requires iOS 17.0 or later.
iPhone Requires iOS 17.0 or later.
Mac Requires macOS 14.0 or later and a Mac with Apple M1 chip or later.
