AndroidBench: Results as Published

Android Bench 2.0 measures whether a coding agent can finish long Android engineering tasks. The page describes 30 private tasks in four streams: creating an app from design mocks, migrating production apps to newer libraries, adding features such as widgets or camera support, and converting Flutter or React Native apps to native Android. Every task runs in a fresh container with an Android emulator. The primary score is the pass rate, the share of runs in which every functional test passes, the visual checks match, and there are no constraint violations.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.