What is Nndeploy used for?

nndeploy is a C++ and Python framework for deploying AI models across desktop, mobile, edge and server hardware using visual workflows and many inference engines. It is a good fit for deploying AI models to edge devices, and running one workflow across desktop and mobile.

Is Nndeploy open source?

Yes. Nndeploy is open source under the Apache-2.0 licence. Its source code is on GitHub at nndeploy/nndeploy and is written mainly in C++.

Yes. Nndeploy is open source, so the software itself is free to use.

Can I self-host Nndeploy?

Yes. Nndeploy can be self-hosted on your own server or infrastructure.

What are some alternatives to Nndeploy?

Similar open-source tools in the AI Infrastructure category include Dify, llama.cpp and AppPlatform. SaaS products in the same category include AI Stats, Ballast and Cloudflare Workers AI.

Is Nndeploy actively maintained?

Yes. The most recent commit to Nndeploy was on 15 August 2026, and the latest release is v3.0.10, published on 4 April 2026. The project has 1.9k stars on GitHub.

Nndeploy: open-source AI infrastructure

About Nndeploy

nndeploy is an AI deployment framework aimed at running AI algorithms on end devices and servers: desktop systems on Windows and macOS, mobile devices on Android and iOS, edge hardware such as NVIDIA Jetson, Ascend and Rockchip boards, and single-machine servers with GPUs. It is built around a visual workflow editor and multi-backend inference.

Users build deployments by dragging nodes onto a canvas with adjustable parameters, can add custom nodes in Python or C++/CUDA, and export workflows as JSON that C++ or Python APIs can run on Linux, Windows, macOS and Android. It integrates a dozen inference frameworks, including ONNX Runtime, TensorRT, OpenVINO, MNN, ncnn, Core ML, AscendCL, RKNN, SNPE, TVM and PyTorch. Performance features include pipeline and task parallelism, zero-copy and memory pooling, and built-in optimized nodes. The library offers more than a hundred nodes and example models such as Qwen language models, Stable Diffusion variants, face swapping and OCR.

nndeploy is licensed under Apache-2.0, has English and Chinese documentation, and runs on hardware you control. For large language and image generation models it positions itself as a visual workflow tool.

Key features

Visual drag-and-drop AI workflows
Custom nodes in Python or C++
Workflow export to JSON
Integration of many inference engines
Parallelism and memory optimizations
Ready-made nodes for LLM, diffusion and OCR

Good fit for

→Deploying AI models to edge devices
→Running one workflow across desktop and mobile

Built with: C++; Python
Tags: ai-deployment; inference; low-code; workflow; cpp; tensorrt; onnxruntime; edge-ai

Nndeploy: questions and answers

What is Nndeploy used for?: nndeploy is a C++ and Python framework for deploying AI models across desktop, mobile, edge and server hardware using visual workflows and many inference engines. It is a good fit for deploying AI models to edge devices, and running one workflow across desktop and mobile.
Is Nndeploy open source?: Yes. Nndeploy is open source under the Apache-2.0 licence. Its source code is on GitHub at nndeploy/nndeploy and is written mainly in C++.
Is Nndeploy free?: Yes. Nndeploy is open source, so the software itself is free to use.
Can I self-host Nndeploy?: Yes. Nndeploy can be self-hosted on your own server or infrastructure.
What are some alternatives to Nndeploy?: Similar open-source tools in the AI Infrastructure category include Dify, llama.cpp and AppPlatform. SaaS products in the same category include AI Stats, Ballast and Cloudflare Workers AI.
Is Nndeploy actively maintained?: Yes. The most recent commit to Nndeploy was on 15 August 2026, and the latest release is v3.0.10, published on 4 April 2026. The project has 1.9k stars on GitHub.