Deep Learning Inference Engine
15.0k 2026-04-28
alibaba/MNN
MNN is a blazing-fast, lightweight deep learning inference engine optimized for high-performance on-device AI and Large Language Models.
Core Features
Blazing-fast and lightweight inference performance.
Optimized for on-device LLMs and Edge AI applications.
Battle-tested across over 30 Alibaba apps and 70 scenarios.
Supports a wide range of popular LLM models like Qwen, Baichuan, and LLAMA.
Cross-platform deployment on mobile, PC, and IoT devices.
Detailed Introduction
MNN is a highly efficient and lightweight deep learning framework that excels in on-device inference and training. Developed and battle-tested by Alibaba, it powers over 30 of their applications, including Taobao and Tmall, across more than 70 diverse scenarios from live broadcasting to search recommendations. Its MNN-LLM solution extends this capability to deploy large language models locally on mobile phones, PCs, and IoT devices, supporting popular models like Qianwen and LLAMA, making advanced AI accessible on edge devices.