NCNN
Neural network inference on mobile devices is the focus of ncnn, a software framework optimized for mobile platforms.
About NCNN
ncnn is a neural network inference framework from Tencent, designed to run trained deep learning models on mobile, embedded, and desktop devices. Its README describes it as having no third-party runtime dependencies and running on CPU and Vulkan GPU backends. The project includes pnnx, a tool that converts PyTorch and ONNX models to the ncnn format. The README states that developers can deploy models on phones, PCs, browsers, and edge devices, and that ncnn is used in several Tencent applications, including QQ, Qzone, WeChat, and Pitu.
Support for each target depends on building the library. The README points to separate build guides for Android, iOS, Linux, Windows, macOS, WebAssembly, HarmonyOS, Raspberry Pi, NVIDIA Jetson, and other embedded targets. The Vulkan GPU backend is enabled through a build option. Pricing information is not stated on the page.
Key features
- Converts PyTorch and ONNX models to the ncnn format using the pnnx tool
- Runs inference on CPU and Vulkan GPU backends
- Provides a C++ API and a Python API for loading and running models
- Has no third-party runtime dependencies, according to the README
- Includes build guides for Android, iOS, Linux, Windows, macOS, and WebAssembly
- Enables Vulkan GPU support through the NCNN_VULKAN CMake build option
Best for: Developers who need to deploy PyTorch or ONNX neural network models on mobile, embedded, or desktop devices.
Good for
- Has no third-party runtime dependencies, according to the README
- Supports both CPU and Vulkan GPU backends
- The README states it is used in several Tencent applications, including QQ and WeChat
Watch out for
- Vulkan GPU support must be enabled at build time with the NCNN_VULKAN option
- Building from source requires platform-specific steps described in separate guides
NCNN at a glance
- Website
- github.com/Tencent/ncnn
- Category
- Development › Developer Tools
- Type
- Open-source project
- Pricing
- Not confirmed. Check the site
- Open source
- Not stated
- Language
- English
- Last checked
- Oct 6, 2026 · Online when checked
- Listed
- Oct 5, 2026
Guides for Developer Tools
NCNN alternatives
Tools and sites you could use instead, ranked by how closely they match.
React Native Vision Camera is a high-performance camera library for React Native apps.
Open source
24 alternatives
Yet-Another-Bench-Script estimates Linux server performance using fio, iperf3, and Geekbench.
Tracy Profiler collects real-time, nanosecond-resolution telemetry and frame performance data.
Audit websites for performance metrics and best practices with Lighthouse, an automated web quality tool.
Open source
24 alternatives
High-performance, high-precision CPU, GPU, and memory profiler for Python.
Open source
24 alternatives
Project Wallace analyzes CSS complexity, specificity, performance, and design tokens.
Popular in Developer Tools
Well-known listings from the same topic.
AppDevToolsappdevtools.comCommand Not Foundcommand-not-found.comTestFlighttestflight.apple.com/join/A8GwCnq8Sourcegraphabout.sourcegraph.comCarbon Calculatorwebsitecarbon.comAwesome CLIgithub.com/umutphp/awesome-cli
NCNN FAQ
How do I convert a PyTorch model for ncnn?
Install pnnx with pip3, then call pnnx.export on the model with a sample input. The export generates a model.ncnn.param file and a model.ncnn.bin file.
Which programming interfaces can I use to run a model?
The README shows examples using the ncnn C++ API and the Python API. Both load the .param and .bin files, pass input to the network, and extract output.
Which platforms have build instructions?
The README links to build guides for Android, iOS, macOS, Linux, Windows, WebAssembly, HarmonyOS, Raspberry Pi, NVIDIA Jetson, and several other embedded targets.
Is something wrong or out of date?
Tell us and an editor will check it.
Own this website?
Claim this listing
Claim the listing and suggest updates yourself. An editor checks every claim first.