Google on Tuesday revealed new Android development tools, a new mobile AI architecture, and an expanded developer community. The announcements accompanied the unveiling of an AI Mode for Google Search at the Google I/O keynote in Mountain View, California.
Gemini 2.5 Pro added to Gemini Studio
Gemini 2.5 Pro is now accessible through both Android Studio’s latest preview release update and Android Studio for Business. Android Studio users can explore agentic AI features with Gemini, tasking AI agents with following specific user journeys through their apps. These journeys can be tracked on local physical or virtual Android devices and will be reflected within the IDE to detailed examination.
SEE: Material 3 Expressive, Gemini AI for Android Auto, and enhanced security controls are coming to Android throughout 2025.
“Although the feature is experimental, the goal is to increase the speed that you can ship high-quality code, while significantly reducing the amount of time you spend manually testing, validating, or reproducing issues,” wrote Android Studio Product Manager Mayank Jain in a blog post.
Other new features for Android Studio enabled by Gemini are:
- App Quality Insights, a panel that suggests explanations for crashes and offers source fixes.
- Experimental AI features, which will now appear in the Studio Labs menu once the features reach stable release.
- Jetpack Compose preview code generation.
- Natural language prompts for changing UI code within the Compose Preview environment.
Gemini also enables users to attach images to prompts, attach project files as context in chats with Gemini, and set preferred rules for Gemini in the Prompt Library.
Google also released a flurry of quality-of-life updates for Android as well, including resizable previews and an Android XR emulator.
Gemma3n AI architecture in early preview for Android and Chrome
Gemma3n, an open AI model built on Google’s latest mobile AI architecture, is now accessible to developers in preview. With tuning from Google DeepMind, Gemma 3n reduces the amount of RAM needed for the AI run on-memory to 2GB or 3GB, depending on parameter size. The company said Gemma3n is 1.5x faster on mobile than Gemma 3 4B. It enables multilingual and multimodal performance. Gemma 3n is available in Google AI Studio and Google AI Edge.
Developer community opens doors to Google and NVIDIA resources
Google and NVIDIA have launched a joint educational portal and forum to encourage developers to prototype projects with both companies. The Google Cloud and NVIDIA community will soon offer AI use credits as incentives for completing designed learning paths.
SEE: More Google I/O news, including details about Search’s new AI Mode.
