Google Updates Android Bench: Gemini 4 Argon Trails Anthropic and OpenAI
Google has updated its Android Bench leaderboard, introducing results for models like Gemini 4 Argon, Claude Sonnet 5.5, and GPT-6.1 Sol. Google's Gemini 4 Argon debuted in fifth place, lagging significantly behind competitors in Android development tasks. This benchmark update highlights the performance gaps and cost-performance trade-offs among frontier AI models in specialized mobile development scenarios. It helps developers choose the most efficient tool by factoring in both task completion rates and API call costs. Claude Sonnet 5.5 claimed the top spot with a 40% pass rate, while OpenAI's GPT-6.1 Sol offered high completion rates at a drastically lower cost of $99.50. Meanwhile, Gemini 4 Argon achieved a 21.3% pass rate and an average invocation cost of $267.70, ranking behind even Claude Fable 5.1.
## BACKGROUND
Android Bench is a specialized framework and benchmark designed by Android Developers to evaluate Large Language Models on mobile codebases, code patch generation, and Android-specific engineering challenges. It evaluates how well AI models handle real-world mobile development tasks and adhere to best practices.