software-mansion/react-native-executorch

High-performance, privacy-first on-device AI inference library for React Native, powered by PyTorch's ExecuTorch runtime

What it solves

It enables developers to run machine learning models directly on mobile devices within React Native applications. This eliminates the need for cloud-based inference, removing network latency, reducing server costs, and ensuring that user data never leaves the device for maximum privacy.

How it works

The library is powered by Meta's ExecuTorch runtime. It provides a two-layer architecture: a high-level layer with ready-to-use task hooks (like useLLMChatSession) for common AI tasks, and a low-level layer that allows developers to build custom pipelines using TypeScript, tensor operations, and worklet threading. It leverages native hardware acceleration via XNNPACK (CPU), Core ML and MLX (Apple Silicon), and Vulkan (Android GPU).

Who it’s for

React Native developers who want to integrate on-device AI capabilities—such as LLMs, computer vision, speech-to-text, and embeddings—into their mobile apps without relying on external APIs.

Highlights

  • Hardware Accelerated: Supports Core ML, MLX, Vulkan, and XNNPACK for optimized performance.
  • Privacy-First: 100% offline execution ensures no data is transmitted to the cloud.
  • Pre-Exported Models: Includes a curated catalog of verified models available via Hugging Face.
  • Flexible Implementation: Offers both simple task-based hooks and low-level primitives for custom .pte model orchestration.

Related

  • Project
    pytorch/executorch

    ExecuTorch is an open‑source PyTorch‑based runtime for deploying AI models on edge devices—from phones to micro‑controllers. It exports models directly from PyTorch, compiles them ahead‑of‑time into a tiny `.pte` file, and runs them on a variety of hardware back‑ends via lightweight C++/Python/Swift/Kotlin APIs. It supports vision, speech, and LLMs (e.g., Llama) and is used in Meta’s production apps.

  • Project
    google-ai-edge/litert-torch

    A Python library for converting PyTorch models into .tflite format, enabling high-performance on-device execution on Android, iOS, and IoT devices.

  • Project
    google-ai-edge/litert-samples

    A collection of official and community samples, model recipes, and deployment tools for LiteRT and LiteRT-LM to enable high-performance on-device ML execution.

  • Project
    margelo/react-native-fast-tflite

    A high-performance TensorFlow Lite library for React Native that enables efficient on-device AI inference using zero-copy ArrayBuffers and GPU acceleration.

  • Project
    google-ai-edge/mediapipe

    An open-source framework for building and deploying customizable on-device machine learning pipelines across mobile, web, desktop, and IoT devices.