Deep Learning Inference Engine
14.9k 2026-04-16
alibaba/MNN
A blazing-fast, lightweight inference engine from Alibaba, powering high-performance on-device LLMs and Edge AI.
Core Features
Blazing-fast, lightweight deep learning inference and training engine.
Optimized for on-device deployment across mobile, PC, and IoT platforms.
Supports a wide range of popular Large Language Models (LLMs) like Qwen, Baichuan, and LLAMA.
Powers over 70 AI scenarios in Alibaba's major applications, demonstrating battle-tested reliability.
Enables multimodal AI capabilities including text, image, and audio processing on edge devices.
Detailed Introduction
MNN is a highly efficient and lightweight deep learning framework developed by Alibaba, offering industry-leading performance for on-device inference and training. It has been extensively integrated into over 30 Alibaba applications, covering more than 70 usage scenarios, and is also deployed on embedded devices like IoT. MNN-LLM, built upon the MNN engine, extends its capabilities to deploy large language models locally on personal platforms such as mobile phones, PCs, and IoT devices, supporting popular LLMs and empowering edge AI applications.