Google updates Android Bench with new large language models
Google has updated Android Bench, its evaluation platform for large language models (LLMs) in Android app development. The update introduces new models and enhanced metrics for assessing LLM performance.
Google has updated Android Bench, its benchmark for evaluating how large language models (LLMs) perform in Android app development tasks. The revised platform now incorporates eight new models, including recent versions of prominent systems.
Launched earlier this year, Android Bench assesses the efficiency of LLM agents across a suite of 100 Android development tasks. The platform aims to assist developers in distinguishing useful code suggestions from less reliable outputs. The update also introduces metrics for cost and efficiency, alongside open-weight models.
The newly added models include 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. Google is inviting developers to run their own tests and submit feedback to guide the future development of Android Bench.
As LLMs become increasingly used for code generation, their performance varies significantly across different development tasks. Android Bench seeks to provide clarity in this area by offering a standardized comparison framework.