# ggml\-org/ggml

ggml: A Simple, Portable, and Efficient C/C\+\+ Tensor Library for Machine Learning

- HackerByte page: <https://hackerbyte.net/en/projects/ggml-a-simple-portable-and-efficient-c-c-tensor-library-for-machine-learning-538180525>
- GitHub: <https://github.com/ggml-org/ggml>
- Programming language: C\+\+

Tensor library for machine learning

## AI-generated summary

ggml is a tensor library designed for machine learning, emphasizing simplicity, portability, and efficiency. It's implemented in plain C/C\+\+ with no external dependencies, supporting a wide range of architectures and backends, including CPU, GPU, and NPU. The library offers various quantization formats and aims for zero memory allocations during runtime.

## Key points

- Implemented in plain C/C\+\+ with minimal dependencies.
- Cross\-platform support for x86, ARM, RISC\-V, and WebAssembly.
- Optimized kernels for x86, ARM, and RISC\-V with SIMD support.
- Supports multiple backends: CPU, GPU, NPU, and browser.
- Offers 2\- to 8\-bit integer quantization, plus MXFP4 and NVFP4 formats.
