Google moves LiteRT into Play services with faster GPU inference and smaller apps
Overview
Google has moved LiteRT, its on-device machine learning runtime, into Google Play services.
Google says GPU inference runs about 1.4x faster than the legacy TensorFlow Lite GPU delegate through the new CompiledModel API. In Google's testing, apps that switched to the Play services runtime were about 8 MB smaller, because the runtime ships once with Play services instead of inside each app.
Developers use it by adding com.google.android.gms:play-services-litert:17.0.0 or a later version. Google says the legacy GPU delegate and custom delegate APIs are deprecated in favor of CompiledModel GPU acceleration and the Compiler Plugin and Dispatch APIs. NNAPI will be removed starting in 2027.
Written by AI from the articles below · updated Oct 28, 5:08 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Google Developers BlogOfficialGoogle moves LiteRT into Play services for faster on-device GPU inference
AIGoogle says LiteRT in Google Play services runs GPU inference about 1.4x faster than the legacy TensorFlow Lite GPU delegate through the new CompiledModel API. In Google's testing, apps that switched to the Play services runtime were about 8 MB smaller because the runtime ships once with Play services rather than inside each app. The post also notes that the legacy GPU delegate and NNAPI are deprecated, with NNAPI to be removed starting in 2027.
Heat Trend
Not enough continuous observations to show a trend yet.