How open-weight models stack up on Android Bench @AndroidDevelopers
How open-weight models stack up on Android Bench  @AndroidDevelopers
Uploaded June 2026 | Updated November 2026, 2 days ago
We put 13 open-weight models through Android Bench to see how they handle real Android development tasks: From large frontier models to ones you can run locally, like Gemma 4; all of them accessible directly on Android Studio with Bring Your Own Model. Check their scores, costs, latency, and token counts in the Android Bench leaderboard to find the best match for your needs.

Resources:
Android Bench website → https://goo.gle/3Q7NB0b
Methodology for benchmarking open-weight models → https://goo.gle/4ovhMen
Open-source dataset and test harness → github.com/android-bench/android-bench
Android Studio Bring Your Own Model →youtube.com/watch?v=gQuUhIfjTyI
Subscribe to Android Developers → https://goo.gle/AndroidDevs

#AndroidDevelopers #AndroidBench

Products Mentioned: Android Bench, Android

Speakers: Zoe Lopez-Latorre
How open-weight models stack up on Android BenchElevating camera experiences across devices with CameraX and on-device intelligenceMoving beyond 2D with custom 3D geometry in the Jetpack XR SDKAndroid Dev Summit: Introducing XREAL AURABuilding for Android XR with OpenXR Spatial ContainersWhat I tried at I/O Connect: Android CLIBuilding a Formula E co-driver on Android with ML Kit and Gemini NanoBuild for Android XR with your preferred 3D engineHow to migrate to Compose with an Android skillThe latest in Android XRHow to make your Android app smarter with AppFunctions APIOn-device intelligence: Optimized performance and quality using Android ML Kit APIs
Android Developers |

How open-weight models stack up on Android Bench

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER