# Install ncnn with Homebrew, dnf, Nix, pacman, zypper

High-performance neural network inference framework. Version 20260526 via Homebrew; verified 2026-06-25. Also installable with nix: nix profile install nixpkgs#ncnn.

## Install

```sh
sudo av install brew:ncnn
```

Additional install commands:

### macOS

- Homebrew (100%):

```sh
brew install ncnn
```

  Evidence: local Homebrew formula metadata

### Linux

- dnf (92%):

```sh
sudo dnf install ncnn
```

  Evidence: Fedora Rawhide package metadata: ncnn from https://dl.fedoraproject.org/pub/fedora/linux/development/rawhide/Everything/x86_64/os/repodata/07190dc5ae9f35ae73866675fed6d95fe6e8d9fe22c9d7cdf85862cb2ed24a4c-primary.xml.zst

- Nix (92%):

```sh
nix profile install nixpkgs#ncnn
```

  Evidence: nixpkgs package indexes: pkgs/by-name/nc/ncnn/package.nix from https://api.github.com/repos/NixOS/nixpkgs/git/trees/master?recursive=1

- pacman (92%):

```sh
sudo pacman -S ncnn
```

  Evidence: Arch Linux sync databases: ncnn from https://geo.mirror.pkgbuild.com/extra/os/x86_64/extra.db.tar.gz

- zypper (92%):

```sh
sudo zypper install ncnn
```

  Evidence: openSUSE Tumbleweed package metadata: ncnn from https://download.opensuse.org/tumbleweed/repo/oss/repodata/50b07339cb64c8ed4091bdbabddadc1ff5737b090e478818a195b40d8a3292861a879139b4a3987c31109699fde9fbf4a716367ddf4eef77da75f96e3193d6ed-primary.xml.zst

## Package facts

- **Package key:** brew:ncnn
- **Package manager:** Homebrew
- **Version:** 20260526
- **Source summary:** High-performance neural network inference framework
- **Homepage:** <https://github.com/Tencent/ncnn>
- **Repository:** <https://github.com/Tencent/ncnn>
- **Last updated:** 2026-06-25T13:37:56+02:00
- **Generated:** 2026-08-03T19:37:03+00:00

## Executables

- caffe2ncnn (alias)
- darknet2ncnn (alias)
- mxnet2ncnn (alias)
- ncnn2int8 (alias)
- ncnn2mem (alias)
- ncnn2table (alias)
- ncnnmerge (alias)
- ncnnoptimize (alias)

## Install behavior

- Bottle: not available

## Freshness

- Page generated: 2026-08-03
- Package-manager version: 20260526
## Project history and usage

ncnn is Tencent's open-source neural-network inference runtime for mobile, embedded, desktop, browser, and edge deployment. Its niche is small, dependency-light inference: the upstream README emphasizes no third-party runtime dependencies, CPU and Vulkan GPU backends, explicit low-memory allocator design, and conversion tools such as pnnx for PyTorch and ONNX workflows.

### Project history

Tencent presentations describe the ncnn project as dating from 2017, with open collaboration and Vulkan acceleration as central design points. That timing put it in the wave of mobile inference runtimes that appeared after deep-learning deployment moved from server GPUs into phones and edge devices.

### Adoption history

Upstream states that ncnn is used in Tencent applications including QQ, Qzone, WeChat, and Pitu. Outside Tencent, its adoption pattern is typical for low-level ML infrastructure: package-manager availability, examples for object detection and vision models, and use by developers who want a C++ runtime without TensorFlow Lite or ONNX Runtime-sized baggage.

### How it is used

Package users usually install ncnn for the converter and optimizer executables as much as for the library. A common flow is export from PyTorch through pnnx, produce .param and .bin model files, run optimization or int8 tooling, and embed the resulting model in a C++ or mobile application using ncnn::Net.

### Why package nerds care

For package maintainers, ncnn is interesting because it is a fast-moving C++/Vulkan project that exposes both libraries and many small CLI tools. It is also a useful example of a Homebrew formula where the command-line surface is mostly build/deployment tooling around a library rather than one end-user command.

### Timeline

- 2017: Tencent presentation material describes ncnn as active since 2017.
- 2020s: README documents pnnx-based PyTorch/ONNX conversion and CPU/Vulkan deployment across mobile, desktop, browser, and edge targets.

### Related projects

- PyTorch, ONNX, pnnx, Vulkan, TensorFlow, Caffe, MXNet, Darknet

### Sources

- <https://formulae.brew.sh/formula/ncnn>
- <https://github.com/Tencent/ncnn>
- <https://www.khronos.org/developers/linkto/ncnn-universal-neural-network-inference-with-vulkan>


## Security Notes

broad file, network, media, or database tool signal.

- **Geiger risk:** blue / medium
- broad file, network, media, or database tool signal

## Other Package-Manager Records

- Nix - ncnn: normalized package name match | nixpkgs package indexes: pkgs/by-name/nc/ncnn/package.nix from https://api.github.com/repos/NixOS/nixpkgs/git/trees/master?recursive=1
- dnf - ncnn - 20250916-5.fc45: normalized package name match | Fedora Rawhide package metadata: ncnn from https://dl.fedoraproject.org/pub/fedora/linux/development/rawhide/Everything/x86_64/os/repodata/07190dc5ae9f35ae73866675fed6d95fe6e8d9fe22c9d7cdf85862cb2ed24a4c-primary.xml.zst | A high-performance neural network inference framework | https://github.com/Tencent/ncnn
- dnf - ncnn-devel - 20250916-5.fc45: normalized package name match | Fedora Rawhide package metadata: ncnn-devel from https://dl.fedoraproject.org/pub/fedora/linux/development/rawhide/Everything/x86_64/os/repodata/07190dc5ae9f35ae73866675fed6d95fe6e8d9fe22c9d7cdf85862cb2ed24a4c-primary.xml.zst | Development files for ncnn | https://github.com/Tencent/ncnn
- pacman - ncnn - 20260526-6: normalized package name match | Arch Linux sync databases: ncnn from https://geo.mirror.pkgbuild.com/extra/os/x86_64/extra.db.tar.gz | High-performance neural network inference framework optimized for the mobile platform | https://github.com/Tencent/ncnn
- zypper - libncnn1 - 20250916-1.10: normalized package name match | openSUSE Tumbleweed package metadata: libncnn1 from https://download.opensuse.org/tumbleweed/repo/oss/repodata/50b07339cb64c8ed4091bdbabddadc1ff5737b090e478818a195b40d8a3292861a879139b4a3987c31109699fde9fbf4a716367ddf4eef77da75f96e3193d6ed-primary.xml.zst | NCNN library | https://github.com/Tencent/ncnn
- zypper - ncnn - 20250916-1.10: normalized package name match | openSUSE Tumbleweed package metadata: ncnn from https://download.opensuse.org/tumbleweed/repo/oss/repodata/50b07339cb64c8ed4091bdbabddadc1ff5737b090e478818a195b40d8a3292861a879139b4a3987c31109699fde9fbf4a716367ddf4eef77da75f96e3193d6ed-primary.xml.zst | A high-performance neural network inference framework | https://github.com/Tencent/ncnn
- zypper - ncnn-devel - 20250916-1.10: normalized package name match | openSUSE Tumbleweed package metadata: ncnn-devel from https://download.opensuse.org/tumbleweed/repo/oss/repodata/50b07339cb64c8ed4091bdbabddadc1ff5737b090e478818a195b40d8a3292861a879139b4a3987c31109699fde9fbf4a716367ddf4eef77da75f96e3193d6ed-primary.xml.zst | Development tools for ncnn | https://github.com/Tencent/ncnn


## Combined YAML source

View the package source record on GitHub. [combined/ncnn.yml](https://github.com/mxcl/pkgdb/blob/main/combined/ncnn.yml)


## Sources

- pkg.so package database
- Geiger risk classifier
- curated package history
- pkgdb category and tag curation
- external package-manager database matches
- cross-ecosystem install command graph