Author Archives:

Gemma 4: The new standard for local agentic intelligence on Android

Posted by Matthew McCullough, VP of Product Management Android Development



Today, we are enhancing Android development with Gemma 4, our latest state-of-the-art open model designed with complex reasoning and autonomous tool-calling capabilities.

Our vision is to enable local agentic AI on Android across the entire software lifecycle, from development to production. Android supports a range of Gemma 4 models, from the most efficient ones running directly on-device in your apps to more powerful ones running on your development machine to help you build apps. We are bringing Gemma 4 to Android developers through two pillars:

  • Local-first Agentic coding: Experience powerful, local AI code assistance with Gemma 4 in Android Studio in your development computer.
  • On-device intelligence: Build intelligent experiences using the ML Kit GenAI Prompt API to run Gemma 4 directly on Android device hardware.

Coding with Gemma 4 in Android Studio

When building Android apps, Android Studio can use Gemma 4 to leverage its state-of-the-art reasoning power and native support for tool use, while keeping the model and inference contained entirely on your local machine.

Gemma 4 was trained on Android development and designed with Agent Mode in mind. This means that when you select Gemma 4 as your local model, you can leverage the full suite of Agent Mode capabilities for a variety of Android development use cases, including refactoring legacy code, building an entire app or new features, and applying fixes iteratively.

Learn more about the possibilities Gemma 4 brings to your app development flow and how to get started.

Prototyping with Gemma 4 on-device

Since the introduction of Gemini Nano as the foundation model on Android, it has become available on over 140 million devices. Gemma 4 is the base model for the next generation of Gemini Nano (Gemini Nano 4) that is optimized for performance and quality on Android devices. This model is up to 4x faster than the previous version and uses up to 60% less battery.

To make it as easy as possible to preview and prototype with Gemma 4 E2B and E4B models directly on AICore-supported devices, we’re launching the AICore Developer Preview. While we continue to expand the ML Kit GenAI Prompt API surface to unlock additional advanced capabilities of the model, you can already start exploring new use cases with Gemma 4 using the Prompt API.

Prepare your apps for the launch of the Gemini Nano 4 on the new flagship Android devices later this year by prototyping with Gemma 4 today. Read about the upcoming features and deep dive into AICore Developer Preview and its Gemma 4 support here.

Local agentic intelligence with Gemma 4

Running Gemma 4 locally, you can leverage its advanced reasoning and tool-calling capabilities in your entire workflow, from developing with the AI coding assistant in Android Studio to shipping intelligent features in your app with ML Kit GenAI Prompt API. This local-first approach, available under Gemma’s open Apache license, provides an alternative for developers to innovate in a privacy-centric and cost effective manner. We're updating Android Bench to include Gemma 4 and other open models, providing the quantified data you need to navigate performance trade-offs and select the best model for your use case.

We can’t wait to see what you build!

Android Studio supports Gemma 4: our most capable local model for agentic coding

Posted by Matthew Warner, Google Product Manager


Every developer's AI workflow and needs are unique, and it's important to be able to choose how AI helps your development. In January, we introduced the ability to choose any local or remote AI model to power AI functionality in Android Studio, and today, we're announcing the availability of Gemma 4 for AI coding assistance in Android Studio. This new local model trained on Android development provides the best of both worlds: the privacy and cost-efficiency of on-device processing alongside state-of-the-art reasoning and tool-calling capabilities.

AI assistance, locally delivered

By running locally on your machine, Gemma 4 gives you AI code assistance that doesn't require an internet connection or an API key for its core operations. Key benefits include:

  • Privacy and security: Your code stays on your machine. Gemma 4 processes all Agent Mode requests locally, making it an ideal choice for developers working with data privacy requirements or in secure corporate environments.
  • Cost efficiency: Run complex agentic workflows without worrying about hitting quotas. Gemma 4 is optimized to run efficiently on modern development hardware, utilizing local GPU and RAM to provide snappy, responsive assistance.
  • Offline availability: Use the agent to write code even when you don’t have an internet connection.
  • State-of-the-art reasoning: Gemma 4 delivers best-in-class reasoning, capable of complex multi-step coding tasks in Agent Mode.

Powerful agentic coding

Gemma 4 was trained for Android development with agentic tool calling capabilities. When you select Gemma 4 as your local model, you can leverage Agent Mode for a variety of development use cases, such as:

  • Designing new features: Developers can ask the agent to build a new feature or an entire app with commands like “build a calculator app” and the agent will not only generate the UI code but will use Android best practices like writing in Kotlin and using Jetpack Compose.
  • Refactoring: You can give high-level commands such as "Extract all hardcoded strings and migrate them to strings.xml." The agent will scan your codebase, identify instances requiring changes, and apply the edits across multiple files simultaneously.
  • Bug fixing and build resolution: If a project fails to build or has persistent lint errors, you can prompt the agent to "Build my project and fix any errors." The agent will navigate to the offending code and iteratively apply fixes until the build is successful.



Recommended hardware requirements

The 26B MoE is recommended for Android app developers using a machine with the minimum hardware requirements. Total RAM needed includes both Android Studio and Gemma.

Model Total RAM needed Storage needed
Gemma E2B 8GB 2 GB
Gemma E4B 12 GB 4 GB
Gemma 26B MoE 24 GB 17 GB

Get started

To get started, ensure you have the latest version of Android Studio installed.
  1. Install an LLM provider, such as LM Studio or Ollama, on your local computer.
  2. In Settings > Tools > AI > Model Providers add your LM Studio or Ollama instance.
  1. Download the Gemma 4 model from Ollama or LM Studio. Refer to hardware requirements for model size selection.
  2. In Agent Mode, select Gemma 4 as your active model.

For a detailed walkthrough on configuration, check out the official documentation on how to use a local model.

We are excited to see how Gemma 4 enables more private, secure, and powerful development workflows. As always, your feedback is essential as we continue to refine the AI experience in Android Studio. If you find a bug or issue, please file an issue. Also you can be part of our vibrant Android developer community on LinkedIn, YouTube, or X. Happy coding!

Chrome for Android Update

 Hello Everyone! We've just released Chrome 147 (147.0.7727.49) for Android to a small percentage of users. It'll become available on Google Play over the next few days. You can find more details about early Stable releases here.

This release includes stability and performance improvements. You can see a full list of the changes in the Git log. If you find a new issue, please let us know by filing a bug.

Harry Souders
Google Chrome

Get your Wear OS apps ready for the 64-bit requirement

Posted by Michael Stillwell, Developer Relations Engineer and Dimitris Kosmidis, Product Manager, Wear OS

64-bit architectures provide performance improvements and a foundation for future innovation, delivering faster and richer experiences for your users. We’ve supported 64-bit CPUs since Android 5. This aligns Wear OS with recent updates for Google TV and other form factors, building on the 64-bit requirement first introduced for mobile in 2019.

Today, we are extending this 64-bit requirement to Wear OS. This blog provides guidance to help you prepare your apps to meet these new requirements.

The 64-bit requirement: timeline for Wear OS developers

Starting September 15, 2026:

  • All new apps and app updates that include native code will be required to provide 64-bit versions in addition to 32-bit versions when publishing to Google Play.
  • Google Play will start blocking the upload of non-compliant apps to the Play Console.

We are not making changes to our policy on 32-bit support, and Google Play will continue to deliver apps to existing 32-bit devices.

The vast majority of Wear OS developers has already made this shift, with 64-bit compliant apps already available. For the remaining apps, we expect the effort to be small.

Preparing for the 64-bit requirement

Many apps are written entirely in non-native code (i.e. Kotlin or Java) and do not need any code changes. However, it is important to note that even if you do not write native code yourself, a dependency or SDK could be introducing it into your app, so you still need to check whether your app includes native code.

Assess your app

  • Inspect your APK or app bundle for native code using the APK Analyzer in Android Studio.
  • Look for .so files within the lib folder. For ARM devices, 32-bit libraries are located in lib/armeabi-v7a, while the 64-bit equivalent is lib/arm64-v8a.
  • Ensure parity: The goal is to ensure that your app runs correctly in a 64-bit-only environment. While specific configurations may vary, for most apps this means that for each native 32-bit architecture you support, you should include the corresponding 64-bit architecture by providing the relevant .so files for both ABIs.
  • Upgrade SDKs: If you only have 32-bit versions of a third-party library or SDK, reach out to the provider for a 64-bit compliant version.

How to test 64-bit compatibility

The 64-bit version of your app should offer the same quality and feature set as the 32-bit version. The Wear OS Android Emulator can be used to verify that your app behaves and performs as expected in a 64-bit environment.

Note: Since Wear OS apps are required to target Wear OS 4 or higher to be submitted to Google Play, you are likely already testing on these newer, 64-bit only images.

When testing, pay attention to native code loaders such as SoLoader or older versions of OpenSSL, which may require updates to function correctly on 64-bit only hardware.

Next steps

We are announcing this requirement now to give developers a six-month window to bring their apps into compliance before enforcement begins in September 2026. For more detailed guidance on the transition, please refer to our in-depth documentation on supporting 64-bit architectures.

This transition marks an exciting step for the future of Wear OS and the benefits that 64-bit compatibility will bring to the ecosystem.

Upcoming change to default setting for downloading Google Meet recordings

Currently, Google Meet video recordings do not allow viewers to download or copy them by default unless the recording owner explicitly allows it. As a result, the "Ask Gemini" functionality within the Drive viewer is also disabled by default for viewers who aren’t file owners.

Starting April 30, 2026, we will change this default for new recordings. From that date forward, recording owners will need to manually restrict this setting for individual recordings if they do not want viewers to be able to download or copy them. This change applies only to future recordings and will not impact existing files.

If you want to keep downloads disabled by default, you must uncheck "Let Users download and copy Meet Recordings" in the Admin console before April 30, 2026. As a reminder, this will restrict Ask Gemini for viewers unless the recording owner takes action to allow downloads.

Getting started

  • Admins: Admins can manage this via the new "Meet video settings > Let Users download and copy Meet Recordings" control at the domain, OU, or group level. The new default is to allow users to download and copy Meet Recordings. Admins can change this default at any time. Visit the Help Center to learn more.
  • End users: Recording owners retain the ability to manually restrict downloading and copying for individual recordings through the file sharing settings. Visit the Help Center to learn more.

Rollout pace

Availability

  • Business: Business Plus and Business Standard
  • Enterprise: Enterprise Essentials, Enterprise Plus, Enterprise Standard, and Enterprise Starter
  • Education: Education Plus and the Teaching and Learning Upgrade

Resources

Early Stable Update for Desktop

  The Stable channel has been updated to 147.0.7727.49/.50 for Windows and Mac as part of our early stable release to a small percentage of users. A full list of changes in this build is available in the log.

You can find more details about early Stable releases here.

Interested in switching release channels?  Find out how here. If you find a new issue, please let us know by filing a bug. The community help forum is also a great place to reach out for help or learn about common issues.


Krishna Govind

Google Chrome

Developer’s Guide to Building ADK Agents with Skills

The Agent Development Kit (ADK) SkillToolset introduces a "progressive disclosure" architecture that allows AI agents to load domain expertise on demand, reducing token usage by up to 90% compared to traditional monolithic prompts. Through four distinct patterns—ranging from simple inline checklists to "skill factories" where agents write their own code—the system enables agents to dynamically expand their capabilities at runtime using the universal agentskills.io specification. This modular approach ensures that complex instructions and external resources are only accessed when relevant, creating a scalable and self-extending framework for modern AI development.