TensorFlow Lite

TensorFlow Lite (TFLite): The lightweight, cross-platform framework for low-latency machine learning inference on mobile, embedded, and IoT devices.

TensorFlow Lite is Google's open-source, optimized solution for deploying machine learning models to edge devices: this is on-device AI. It converts trained TensorFlow models into the efficient `.tflite` format, enabling low-latency inference on resource-constrained platforms (Android, iOS, embedded Linux, microcontrollers). The framework leverages hardware acceleration (e.g., GPU, NNAPI) and techniques like model quantization to ensure a small binary size and high performance. TFLite is a critical tool for running real-time ML tasks—like object detection or image classification—offline, directly on over 4 billion devices globally.

https://ai.google.dev/edge/litert