tencent/ncnn is a high-performance neural network inference framework optimized for mobile platforms. The purpose of this repository is to make inference (the stage of running an already-trained model) happen on phones as well. It is written in C++, and the latest tag among its GitHub releases is the date-form version 20261020. Since the release record contains no changelog, there is no way to know what this tag changed.
The README describes ncnn as a framework optimized for mobile, embedded, and desktop deployment. The key points are that it has no third-party runtime dependencies and that it offers both CPU and Vulkan GPU backends (the path for running a model on the GPU). With pnnx, PyTorch and ONNX models can be converted into ncnn models, and the deployment targets are broad: phones, PCs, browsers, and edge devices. The GitHub star count is 23985, and the repository was created on June 30, 2017.
There are, however, points where judgment is blocked. The release 20261020 carries no performance figures or benchmarks in the available material. The license is classified only as NOASSERTION, so its specific terms cannot be confirmed. The research record's score is out of 10, with only momentum given at 3.0, while the freshness, technology, and quality items were not computed. It is close to an option worth checking when evaluating mobile and edge device deployment, but from this material alone it is hard to judge which devices and models this repository runs on, and at what level.