Google has released Android Studio Quail 4 to the stable channel, adding built-in Android development skills, native Gemma 4 integration and several updates to the IDE’s agentic coding tools.
The release is the final stable version in the Android Studio Quail series and expands Google’s use of AI inside its Android development environment, particularly for workflows that require platform-specific knowledge or local AI processing.
One of the main additions is a collection of 23 curated Android skills built directly into Android Studio. Google says general-purpose large language models can produce outdated or incorrect code when dealing with changing Android APIs, migrations and complex configuration requirements, prompting the company to integrate Android-specific instructions directly into the IDE.
The skills follow the open-standard agent skills specification and are designed to provide AI agents with instructions for specific Android development workflows. When a developer sends a prompt to the Android Studio agent, the IDE evaluates the request against metadata from the installed skills and can automatically invoke the relevant one.
Among the 23 bundled options are skills covering Android Gradle Plugin 9 upgrades, Android Profiler, Navigation 3 and adaptive app development. Developers can also create custom skills for their own workflows or disable the bundled skills through Android Studio settings.
Google is extending the same system beyond the IDE through Android CLI. Developers using compatible command-line AI tools can install the Android skills through Android CLI, allowing the platform-specific instructions to be incorporated into other agentic development environments.
Quail 4 also introduces native Gemma 4 integration, allowing developers to use Google’s open model for AI coding tasks directly on their development machine. Unlike earlier setups that could require external local model providers, Android Studio can now handle downloading, verifying and updating Gemma 4 model weights itself.
Developers can select Gemma through the Agent model selector or manage the model through Android Studio’s AI settings. Google has also bundled a lightweight inference engine into the IDE to run the models locally.
Because the model operates on the developer’s machine, source code does not need to leave the local environment, and Gemma 4 can continue operating without an internet connection. Local use also avoids token quota limits associated with cloud-hosted models.
Gemma 4’s agentic tool-calling capabilities allow the Android Studio agent to carry out more complex tasks such as multi-file refactoring while running offline. Google says the smallest supported models can run on machines with 12GB of RAM, although systems with at least 32GB of RAM are recommended for the best experience.
The native integration builds on Google’s earlier introduction of Gemma 4 support for Android development. The model was designed for Android-related coding and agentic workflows, including feature development, refactoring, bug fixing and build resolution.
Android Studio Quail 4 also updates the interface for developers running multiple AI agents simultaneously. Code symbols referenced in agent responses — including classes, methods, functions and file paths — can now appear as clickable links that take developers directly to the corresponding code.
The Recent Chats panel has also gained real-time status indicators for parallel agents. Developers can see when an agent is actively working, waiting for additional input or has completed a background task.
For multi-step coding jobs, Google has consolidated the previous Task and Walkthrough outputs into a Summary of Changes tab. The interface provides developers with a centralized view of modifications before they are applied. Reasoning models can also display their step-by-step processes inside collapsible sections to reduce the amount of information shown by default.
Google continues to provide a default Gemini model within Android Studio, while Quail 4 also expands the options for developers requiring additional AI capacity. Developers can connect a Google AI Studio API key for access to newer Gemini models or use API keys from other model providers, including Anthropic and OpenAI.
Google AI Pro and Ultra subscribers can use their Google accounts to access additional capacity and higher rate limits, while organizations with Gemini Enterprise can use its AI capabilities within Android Studio. Enterprise access is currently rolling out to selected organizations through the latest Canary version.
The Quail series has progressively expanded Android Studio’s agentic development capabilities. Previous releases introduced parallel AI conversations, native LeakCanary integration, Planning Mode and an MCP Marketplace for connecting agents with external developer tools and data sources.
With Quail 4 now stable, Google is bringing those features together with built-in Android domain knowledge and a locally managed AI model, giving developers the option to use cloud-based AI services, local Gemma 4 models or other supported model providers within the same development environment.
Comments
Loading…