Google introduces Android Bench, ranking AI models for app development
Technology
Google has rolled out Android Bench, a new tool that officially ranks AI models used in building Android apps.
Think of it as a leaderboard for large language models: right now, Gemini 3.1 Pro is leading the pack, with Claude Opus 4.6 and GPT-5.2-Codex close behind.
How Android bench tests AI models
Android Bench tests these AI models on real tasks developers care about, like wearable networking or updating to the latest Jetpack Compose features, using code from public GitHub projects.
Google is making everything transparent (the methods and data are open on GitHub) so developers can trust the results and see exactly how each model stacks up.