Android Bench, an LLM leaderboard for Android development tasks, was introduced in March to promote transparency and model improvements. Based on feedback, the benchmark now evaluates open-weight models and includes cost and efficiency dimensions. The July release adopts the Harbor framework, an updated benchmarking agent, for more rigorous model evaluations. This update necessitated re-running all models, causing a minor shift in scores, though historical data remains accessible. Eight new models, including Claude Fable 5, Claude Sonnet 5, and GLM 5.2, have been added to the leaderboard. Claude Fable 5 currently leads with an 84.5 score, while GLM 5.2 tops the open-weight models. The Android developer community can now contribute to Android Bench by submitting new development tasks or running and sharing benchmark evaluations. This initiative aims to create a benchmark that truly reflects the diverse challenges faced by global Android developers. The goal is to continuously update Android Bench to ensure AI assistance remains cutting-edge, smarter, and more effective. Developers are encouraged to engage with the GitHub repository and Harbor Hub for contributions and to review the updated leaderboard and methodology.