We've all been there—trying to figure out which AI actually writes working Android code instead of just sounding confident. To make that choice easier, Google just updated Android Bench, its tool for measuring AI agent performance across 100 real-world development tasks. They've introduced a more user-friendly framework and started tracking cost and efficiency, while finally bringing open-weight models into the mix.

The updated leaderboard now features eight new heavy hitters: Claude Fable 5, Claude Sonnet 5, Claude Opus 4.8, GLM 5.2, Kimi K2.7 Code, MiniMax M3, Qwen 3.7 Plus, and Qwen 3.7 Max. If you're a developer, you can jump in, run your own tests, and send feedback. It's a great way to stop guessing and start using the AI that actually gets the job done.