Google's Android Bench adds Claude Sonnet 5.5, GPT-6.1 Sol and Gemini 4 Argon, with Claude on top
Google updated its Android Bench coding leaderboard on October 26, adding newer models and showing the reasoning level used for each run. Claude Sonnet 5.5 leads with a 40% pass rate, ahead of Claude Opus 5.5 and GPT-6.1 Sol, while Google's own Gemini 4 Argon sits sixth at about 21%. Confidence intervals across the top six overlap widely, and the board also lists cost per run.
Why it matters
A vendor-run benchmark that ranks its own model sixth is a useful, cost-aware signal for picking an Android coding agent.