Side-by-side comparison

Open deep learning compiler that optimizes models for diverse CPUs, GPUs and accelerators, including edge targets.

Google's on-device ML runtime formerly called TensorFlow Lite, for Android, iOS, embedded Linux and accelerators.

LiteRT for Microcontrollers: tiny interpreter running quantized models in kilobytes of RAM; the classic TinyML runtime.

PyTorch's on-device runtime (1.0 GA in 2025) for phones, wearables and MCUs with Arm Ethos-U and NPU backends.