Meta releases PyTorch inference framework for edge devices

Meta's PyTorch team released ExecuTorch 1.0, an open-source inference framework that enables direct deployment of PyTorch models to edge devices without conversion or rewriting. The framework supports iOS, Android, embedded systems, and PCs with CPU, GPU, and NPU acceleration. It allows developers to deploy any PyTorch model—including LLMs, vision-language models, and image processing—directly to devices, enabling faster responses, improved privacy through local data processing, and offline functionality. ExecuTorch already powers AI features in Instagram, WhatsApp, Messenger, and Facebook.