Google Quietly Launches AI Edge Gallery: Revolutionary Breakthrough for Running AI Models on Smartphones Offline

AIbase基地

Published inAI News · 5 min read · Jun 4, 2025

Google has quietly launched an experimental app called Google AI Edge Gallery, marking a significant step forward in edge computing and privacy-first AI deployment. This application allows users to run open-source AI models from the Hugging Face platform directly on their smartphones without an internet connection, enabling features such as image generation, text processing, code editing, and more.

A Milestone for Localized AI

Google AI Edge Gallery utilizes Google's LiteRT (Lightweight Runtime) technology to run AI models on device hardware locally, significantly enhancing data privacy and processing speed. The app supports multiple tasks, including "AI Chat" for conversations, "Ask Image" for image analysis, and "Prompt Lab" for text summarization and single-turn tasks. Users can browse and select suitable models through an intuitive interface; model sizes range from lightweight versions at 500MB to advanced editions up to 4GB, catering to different hardware performance requirements.

The app currently supports Android devices, with an iOS version expected soon. As an experimental Alpha release, AI Edge Gallery is distributed via GitHub, requiring users to download the APK file and enable "unknown sources" installation. Although it’s an experimental product, testing indicates its performance rivals early cloud-based models (like GPT-3.5), particularly excelling on devices with powerful hardware.

Core Features

Fully Offline Operation: All AI processing occurs locally on the device, eliminating the need for cloud connectivity and ensuring data privacy. Hugging Face Integration: Supports downloading various open-source AI models from the Hugging Face platform, including Google's self-developed Gemma3n model. Diverse Task Support: Covers functions like image generation, question answering, code generation and editing, making it suitable for developers and tech enthusiasts. Open Source and Flexibility: Licensed under Apache2.0, allowing both commercial and non-commercial use, encouraging feedback from the developer community for optimization.

Industry Impact and Prospects

The launch of AI Edge Gallery reflects Google's strategic positioning in the field of edge AI, aiming to promote the popularity of device-side AI through open-source tools and infrastructure. Unlike traditional cloud-based AI, localized processing not only reduces network dependency but also provides practical solutions for areas without stable internet connections. Industry observers note that this move may redefine the way AI applications are deployed, turning smartphones into part of a distributed AI network.

Despite being constrained by device hardware and model size, Google's Gemma3n models have achieved a balance between memory usage and performance, scoring as high as 1293 points on the LMArena text task leaderboard, demonstrating its competitiveness in lightweight models. In the future, with advancements in hardware and the release of the iOS version, AI Edge Gallery is expected to further expand its influence.

Google AI Edge Gallery is now available for free download via GitHub, with Google encouraging the developer community to participate in testing and provide feedback to optimize the user experience.

Address: https://github.com/google-ai-edge/gallery

Hugging Face Launches Free MCP Tutorial! Master AI Context Protocols in One Day!

Hugging Face once again brings惊喜 to global AI developers by officially releasing the free online course for Model Context Protocol (MCP). As a standardized protocol connecting large language models (LLMs) with external data and tools, MCP is becoming the core technology for AI Agent development. This course covers the foundation of MCP protocol to practical development applications, with concise and practical content aimed at helping developers quickly get started and build efficient AI context interaction systems. AIbase deeply analyzes this...

Open-Source Revolution! Step1X-Edit Lands on Hugging Face, Generating Images with Natural Language, Rivaling GPT-4o!

Step1X-Edit, a groundbreaking open-source AI model, has arrived on Hugging Face. This powerful tool allows users to create images using natural language descriptions, demonstrating performance comparable to GPT-4o. This release marks a significant advancement in accessible AI image generation technology.

Hugging Face's Top Model Leaderboard Unveiled: AI Innovation Continues to Heat Up

Hugging Face recently released its top model leaderboard for the second week of April 2025, encompassing various modalities such as text generation, image generation, and video generation. This highlights the rapid iteration and diverse applications of AI technology. According to AIbase, the models featured in this week's leaderboard not only showcase the innovative vitality of the open-source community but also reflect technical trends such as low-precision training and multi-modal generation. Below is an analysis of the leaderboard highlights, with professional insights provided by the AIbase editorial team. Text Generation Models: Balancing Efficiency and Specialization

Hugging Face Acquires Pollen Robotics, Ushering in a New Era for Robotics

On April 15th, Hugging Face, the renowned open-source large language model platform, announced its acquisition of Pollen Robotics, marking its official entry into the physical robotics field. While specific transaction terms remain undisclosed, the acquisition will bring approximately 20 Pollen Robotics employees to Hugging Face. This represents the company's largest personnel acquisition to date, signifying its ambition in expanding its business areas. Hugging Face's co-founder...

Hugging Face, Prominent Open-Source AI Platform, Acquires Pollen Robotics to Enter Robotics Market

Hugging Face, a leading AI development platform, recently announced the acquisition of Pollen Robotics, a French humanoid robotics startup, marking its strategic foray into the robotics sector. While the financial details of the deal remain undisclosed, the acquisition has generated significant attention. Founded in 2016 by engineers Matthieu Lapeyre and Pierre Rouanet, Pollen Robotics' flagship product, Reachy2, is an advanced humanoid robot already utilized at Cornell...

Hugging Face Acquires Pollen Robotics to Accelerate Open-Source Robotics

Hugging Face, the AI development platform, has announced the acquisition of French robotics startup Pollen Robotics for an undisclosed sum. This marks Hugging Face's first foray into hardware and aims to promote the global adoption and development of open-source robotics. Pollen Robotics, founded in 2016 and based in Bordeaux, France, is known for its open-source humanoid robot, Reachy2. Priced at approximately $70,000, Reachy2 has been adopted by institutions such as Cornell University.

Hugging Face Adds Handy Feature: One-Click Check for Compatible Models

Hugging Face, a leading open-source AI community platform, has launched a highly anticipated new feature: users can quickly see which machine learning models their computer hardware can run via platform settings. Users simply add their hardware information, such as GPU model, to their Hugging Face profile settings (located at the top right corner: Profile Icon > Settings > Local Apps and Hardware).