CoolingProducts
Google's LiteRT in Play services adds GPU inference, about 1.4x faster than the old delegate
Google says LiteRT in Google Play services now offers GPU inference built on ML Drift through the new CompiledModel API, about 1.4x faster than the legacy GPU delegate. Apps that switch are about 8 MB smaller in Google's testing. It needs play-services-litert 17.0.0 or later, and NNAPI will be removed starting in 2027.
- HEAT
- OUTLETS
- 0
- FIRST SEEN
- LAST SIGNAL
Primary source
Google for Developersdevelopers.googleblog.com/smaller-apps-and-faster-gpu-inference-with-litert-in-google-play-services/Coverage · 1 article, oldest first
Wednesday, October 28, 2026
Google for DevelopersPrimary
Smaller apps and faster GPU inference with LiteRT in Google Play services