llama.cpp
Inference in portable C++, from laptops to single-board machines
What is llama.cpp?
The quantisation work here is what put usable models on hardware without a datacentre GPU, and much of the local-model ecosystem is built on top of it. Working with it directly means dealing with build flags and hardware specifics that higher-level wrappers hide from you.
What can you do with llama.cpp?
- Run a model in one command — llama cli -hf ggml-org/Qwen3.5-0.8B-GGUF pulls the GGUF straight from Hugging Face and drops you into an interactive session, with no Python environment involved.
- Serve an OpenAI-compatible endpoint — llama serve -hf <repo> starts a REST server that existing OpenAI clients can be pointed at, and ships a built-in web UI for trying the model.
- Fit models larger than VRAM — Integer quantisation from 1.5-bit through 8-bit, combined with CPU+GPU hybrid inference, lets you offload only part of a model to the GPU and keep the rest in system RAM.
- Target the hardware you have — One source tree covers Metal and ARM NEON on Apple silicon, CUDA on NVIDIA, HIP on AMD, MUSA on Moore Threads, plus Vulkan, SYCL, OpenCL, WebGPU and CANN backends.
- Constrain output with GBNF grammars — A GBNF grammar file forces generation to match a syntax you define, which is how you get reliably structured output out of a small local model.
Before you choose llama.cpp
- Check that the model you plan to serve reuses its earlier work between turns. There is an open report of the server reprocessing the whole prompt each time, which both costs the work again and loses what came before.Reported in#21831
Star history
15 Aug to 28 Aug · +2.1k
Frequently asked questions
Is llama.cpp free for commercial use?
llama.cpp is released under the MIT licence — OSI-approved open source, which permits commercial use.
How can llama.cpp be deployed?
llama.cpp is available as Runs locally / Self-hosted.
Documentation
Reproduced from the ggml-org/llama.cpp README, published under MIT. Read the original ↗
llama.cpp
LLM inference in C/C++
manifesto / ggml / ops / maintainer PRs / compile times / lib llama API / llama-server REST API
Quick start
A few options to get llama.cpp installed on your machine:
- Visit https://llama.app and follow the instructions
- Run with Docker - see our Docker documentation
- Download pre-built binaries from the releases page
- Build from source by cloning this repository - check out our build guide
Once installed:
# Download and run a model directly from Hugging Face
llama cli -hf ggml-org/Qwen3.5-0.8B-GGUF
# Launch OpenAI-compatible API server
llama serve -hf ggml-org/Qwen3.5-0.8B-GGUFDescription
The main goal of llama.cpp is to enable LLM (and VLM) inference with minimal setup and state-of-the-art performance on
a wide range of hardware - locally and in the cloud.
- Plain C/C++ implementation without any dependencies
- Apple silicon is a first-class citizen - optimized via ARM NEON, Accelerate and Metal frameworks
- AVX, AVX2, AVX512 and AMX support for x86 architectures
- RVV, ZVFH, ZFH, ZICBOP and ZIHINTPAUSE support for RISC-V architectures
- 1.5-bit, 2-bit, 3-bit, 4-bit, 5-bit, 6-bit, and 8-bit integer quantization for faster inference and reduced memory use
- Custom CUDA kernels for running LLMs on NVIDIA GPUs (support for AMD GPUs via HIP and Moore Threads GPUs via MUSA)
- Vulkan and SYCL backend support
- CPU+GPU hybrid inference to partially accelerate models larger than the total VRAM capacity
The llama.cpp project is build on top of the ggml library.
Supported backends
| Backend | Target devices |
|---|---|
| BLAS | All |
| BLIS | All |
| CANN | Ascend NPU |
| CUDA | Nvidia GPU |
| HIP | AMD GPU |
| Hexagon [In Progress] | Snapdragon |
| IBM zDNN | IBM Z & LinuxONE |
| MUSA | Moore Threads GPU |
| Metal | Apple Silicon |
| OpenCL | Adreno GPU |
| OpenVINO [In Progress] | Intel CPUs, GPUs, and NPUs |
| RPC | All |
| SYCL | Intel GPU |
| VirtGPU | VirtGPU APIR |
| Vulkan | GPU |
| WebGPU | All |
| ZenDNN | AMD CPU |
Documentation
Tools
Development
Other options in Model Serving
Compare side by side:llama.cpp vs Ollama
OllamaModel ServingPulls a quantised model — one compressed to run in less memory — and starts a local HTTP API in front of it, so running a model is a single command. The obvious starting point for building against a model on your own hardware. For serving many people at once, a server built for throughput will do considerably better.
vLLMModel ServingPaged attention and continuous batching let one GPU serve far more concurrent requests than a naive loop, which matters because agents are unusually chatty. The OpenAI-compatible endpoint means most agent frameworks point at it with a base-URL change. Expect real operational work around GPU memory sizing and model loading times.
LiteLLMModel ServingThe library half lets one function call reach any supported provider in the OpenAI request and response shape. The gateway half is the reason most teams arrive: a self-hosted server that issues virtual keys with their own budgets and rate limits, falls back when a provider fails, spreads load across deployments, and records cost per key. The price is an extra component in the request path that you now operate, and code under the enterprise directory is not covered by the MIT grant.