Jul 12, 2025 10:00:00

MLCommons has released a benchmark app called 'MLPerf Mobile' that can measure the AI processing performance of smartphones, so I tried using it.

MLCommons , an industry group that collects benchmark results for AI infrastructure, has released the benchmark app ' MLPerf Mobile ' for Android smartphones on Google Play. Since it seems that benchmark apps that were previously only distributed as APK files are now easily available, I actually installed it and ran the benchmark.

MLCommons Launches MLPerf Mobile on Google Play Store - MLCommons
https://mlcommons.org/2025/07/mlperfmobile-android/

To install MLPerf Mobile, first tap the link below to open Google Play.

MLPerf Mobile - Apps on Google Play
https://play.google.com/store/apps/details?id=org.mlcommons.android.mlperfbench

Tap 'Install' to install.

When you start MLPerf Mobile, the following screen will be displayed, so tap 'GO'.

The first time you use it, you will be notified that you need to download model data for testing, so tap 'Download'.

The download will then begin.

You can check the download progress at the bottom of the screen.

Once the download is complete, tap the arrow button in the upper left corner of the screen.

Tap 'GO' again to start the benchmark. Please note that there is a waiting time of a few seconds between tapping 'GO' and the benchmark starting. When I tried it on the Pixel 7 and Pixel 9 Pro XL, the app crashed at the start of the 'Stable Diffusion' image generation test on both devices, so this time I disabled only the 'Stable Diffusion' test and then tapped 'GO'.

The benchmark screen looks like this: Between each test there is a waiting period to allow the system to cool down.

Once the test is complete, the results screen will be displayed.

The test results for the Pixel 7 and Pixel 9 Pro XL are as follows. It can be seen that the Pixel 9 Pro XL has higher AI processing performance.

Test Content	Pixel 7	Pixel 9 Pro XL
Image Classification v2	257.90	265.83
Object Detection	249.81	279.32
Image Segmentation v2	237.58	422.83
Language Understanding	90.47	123.30
Super Resolution	30.35	52.84
Stable Diffusion	Unmeasurable	Unmeasurable
Image Classification v2 (Offline)	355.08	568.98

MLPerf Mobile also allows you to check your benchmark results history. By creating a free account, you can also save your benchmark results to the cloud and share them with other devices.