Google's Gemini 4 Argon ranks 5th in Android bench test
Technology
Google's new Gemini 4 Argon AI model just got put to the test for Android coding, and it didn't do so hot.
According to the latest Android Bench results, Gemini 4 landed in fifth place, with only 21.3% of its code passing and an 81.8% completion rate.
Meanwhile, models from Anthropic and OpenAI are leading the pack.
Claude Sonnet 5.5 tops Android Bench
Claude Sonnet 5.5 takes the top spot with a 40% pass rate, while OpenAI's GPT-6.1 Sol scores a solid 29.3% pass rate and even edges out others on completion (82.8%), all at a much lower price, $99.50 versus Claude's $1,574.90.
The Android Bench is a fresh tool for developers to compare how well different AIs handle real coding tasks and costs, making it easier to pick the right coding assistant for their projects.