Google completes 32-bit Coral NPU architecture and releases v1.0 open compiler
AIGoogle Developers Blog says the 32-bit Coral NPU architecture is now complete with the Matrix Execution Unit, configurable for up to 1 TOPS at a 10mW average power target for inference. It also releases the Coral Compiler v1.0, an open-source IREE/MLIR stack that takes JAX, PyTorch, and TensorFlow models to Coral NPU hardware.
AIWhy it matters.
The post explains how the complete 32-bit Coral NPU architecture, its open compiler, and new devkits connect model development to edge silicon, with concrete performance and power figures.
Source details
Google Developers Blog · developers.googleblog.comPublished · added here
Original title: Unifying Edge AI: How Coral NPU’s open standard brings portable, energy-viable intelligence from model to silicon