llama.cpp

license

MIT(OSI)

source

public

governance

ggml-org/Hugging Face(community)

core-gated

ungated

Fully MIT-licensed C/C++ inference engine; entire source public, no proprietary tier.

• https://github.com/ggml-org/llama.cpp/blob/master/LICENSE recorded 2026-06-04

  MIT License text, (c) 2023-2026 The ggml authors

De facto standard local/edge inference engine: GGUF (its native format) is >60% of all quantized models on HF, and llama.cpp powers Ollama, LM Studio, GPT4All, Jan.ai, and HF Inference Endpoints, which collectively serve millions. 115k GitHub stars (fastest OSS-AI project to 100k, Mar 2026) corroborates but is not the basis. No clean PyPI/user count for the C++ engine itself, so usage inferred from ecosystem dominance; placed at 4 (1-10M-equivalent).

• https://github.com/ggml-org/llama.cpp recorded 2026-06-04

  115k stars; README lists Ollama/LM Studio/LocalAI/HF Inference Endpoints as downstream integrations

• https://huggingface.co/docs/hub/gguf-llamacpp recorded 2026-06-04

  GGUF/llama.cpp as the standard local-inference path on the HF Hub

Frontier for portable/CPU/consumer-hardware inference and quantization breadth, but not the datacenter-GPU throughput frontier held by vLLM/SGLang/TensorRT-LLM; scored 4 on a category where C5 is reserved for the throughput-defining server engines.

• https://github.com/ggml-org/llama.cpp recorded 2026-06-04

  README documents backend list, 1.5-8bit quantization, 100+ architectures, multimodal support