No description
Find a file
Carl Philipp Klemm 83eff92560
Some checks failed
ui-publish.yml / CUDA: tiled concat for dim 0 with a transposed src1 (push) Failing after 0s
CUDA: tiled concat for dim 0 with a transposed src1
2026-11-21 11:17:25 +01:00
.devops examples : clean-up old/obsolete stuff (#32184) 2026-11-13 10:33:17 +02:00
.gemini contributing: tighten AI usage policy (#18388) 2025-12-29 16:01:32 +01:00
.github CUDA/HIP: Refactor the use of _sync warp primitives (#31617) 2026-11-15 10:58:36 +01:00
.pi/gg pi : turn .pi/gg into a system prompt addendum (#30762) 2026-10-19 09:11:13 +03:00
app cmake : use PROJECT_SOURCE_DIR instead of CMAKE_SOURCE_DIR (#28771) 2026-09-15 05:26:09 +02:00
benches benches : add Nemotron 3 Nano on DGX Spark (#20652) 2026-03-16 21:50:43 +02:00
ci vulkan: Explicitly include SPIRV-Headers (#31416) 2026-11-14 23:37:14 +01:00
cmake ci: move test models to ggml-org-dev (#32096) 2026-11-11 21:17:16 +01:00
common mtmd: support higgs v3 STT and vibevoice-ASR (#32329) 2026-11-15 21:25:34 +01:00
conversion mtmd: support higgs v3 STT and vibevoice-ASR (#32329) 2026-11-15 21:25:34 +01:00
docs mtmd: add zonos3-mini TTS model (#32020) 2026-11-15 19:44:44 +01:00
examples spec + model: refactor DFlash based speculative decoding methods (#32016) 2026-11-14 09:49:27 +01:00
ggml CUDA: tiled concat for dim 0 with a transposed src1 2026-11-21 11:17:25 +01:00
gguf-py mtmd: support higgs v3 STT and vibevoice-ASR (#32329) 2026-11-15 21:25:34 +01:00
grammars common : implement common_schema internal representation for JSON schemas (#28736) 2026-09-12 16:14:50 -05:00
include spec + model: refactor DFlash based speculative decoding methods (#32016) 2026-11-14 09:49:27 +01:00
licenses refactor : remove libcurl, use OpenSSL when available (#18828) 2026-01-14 18:02:47 +01:00
media ui: update the README banner to the new llama.cpp cover and remove the legacy brand assets (#31789) 2026-11-10 22:43:27 +01:00
models mtmd: add zonos3-mini TTS model (#32020) 2026-11-15 19:44:44 +01:00
pocs cmake : fix build when GGML_CPU=OFF and GGML_CUDA=ON (#29026) 2026-09-18 11:44:03 +03:00
requirements requirements, Fix CI: support Python 3.15 (#32199) 2026-11-13 17:50:03 +01:00
scripts CI/HIP: update fattn symbol names in hip-quality-check (#32331) 2026-11-15 15:53:22 +01:00
skills mtmd: refactoring: remove clip.h, consolidate code into mtmd (#31865) 2026-11-08 15:49:06 +01:00
src model : read qwen3vl output head size from the tensor (#32349) 2026-11-15 19:45:17 +01:00
tests HIP: dedupe experts across tokens in CDNA1 MUL_MAT_ID, Q8_0 fused shared expert 2026-11-21 11:16:59 +01:00
tools mtmd: support higgs v3 STT and vibevoice-ASR (#32329) 2026-11-15 21:25:34 +01:00
vendor vendor: sync stb_image (#32305) 2026-11-15 09:37:30 +01:00
.clang-format fix: apply clang-format to CUDA macros (#16017) 2025-09-16 08:59:19 +02:00
.clang-tidy clang-tidy : disable warning about performance enum size (#16127) 2025-09-22 19:57:46 +02:00
.dockerignore docker : prebuild web UI for s390x build [no release] (#24829) 2026-06-20 05:54:42 -05:00
.ecrc cmake : use PROJECT_SOURCE_DIR instead of CMAKE_SOURCE_DIR (#28771) 2026-09-15 05:26:09 +02:00
.editorconfig vendor: apply deep nested json patch from upstream (#30253) 2026-10-10 00:07:44 +02:00
.flake8 ci : re-enable type/lint-check for skills (#30713) 2026-10-20 01:03:03 +02:00
.gitignore ui: PWA support (#23871) 2026-06-12 15:53:26 +02:00
.gitmodules ggml : remove kompute backend (#14501) 2025-07-03 07:48:32 +03:00
.pre-commit-config.yaml convert.py : add python logging instead of print() (#6511) 2024-05-03 22:36:41 +03:00
AGENTS.md skills: add slice-model (#30677) 2026-10-17 02:11:56 +02:00
AUTHORS llama.cpp : bump version to 0.9.0 (#31646) 2026-11-04 14:06:26 +02:00
build-xcframework.sh metal : add metallib build support for xcframework (#28163) 2026-09-02 07:45:56 +08:00
CLAUDE.md contributing: tighten AI usage policy (#18388) 2025-12-29 16:01:32 +01:00
CMakeLists.txt llama.cpp : bump version to 1.0.0 (#32190) 2026-11-13 11:06:06 +02:00
CMakePresets.json cmake : Add CMake presets for Linux and GCC (#14656) 2025-07-13 08:12:36 +03:00
CODEOWNERS llama_moe_cache: run missed experts on CPU, add tuning script (#30966) 2026-11-08 00:20:58 +08:00
CONTRIBUTING.md docs: Rule of thumb for AI review time [no ci] (#28945) 2026-09-15 14:11:16 +02:00
convert_hf_to_gguf.py feat: external PLE file loading (#31185) 2026-11-09 19:41:01 +01:00
convert_hf_to_gguf_update.py feat: add Inkling 2 support (text + vision) (#32217) 2026-11-15 00:51:31 +01:00
convert_llama_ggml_to_gguf.py ci : switch from pyright to ty (#20826) 2026-03-21 08:54:34 +01:00
convert_lora_to_gguf.py ci : bump ty to 0.0.100 (#32030) 2026-11-10 21:54:14 +01:00
flake.nix nix : update deprecated expressions (#28145) 2026-09-07 15:59:45 +02:00
LICENSE docs : Minor cleanups (#19252) 2026-02-02 08:38:55 +02:00
Makefile make : remove make in favor of CMake (#15449) 2025-08-20 13:31:16 +03:00
mypy.ini convert : partially revert PR #4818 (#5041) 2024-01-20 18:14:18 -05:00
pyproject.toml requirements, Fix CI: support Python 3.15 (#32199) 2026-11-13 17:50:03 +01:00
pyrightconfig.json ci : switch from pyright to ty (#20826) 2026-03-21 08:54:34 +01:00
README.md docs: move per-model TTS to /docs, refer it from other pages (#32304) 2026-11-15 18:35:27 +01:00
requirements.txt tool-call: fix Qwen 2.5 Coder support, add micro benchmarks, support trigger patterns for lazy grammars (#12034) 2025-03-05 13:05:13 +00:00
SECURITY.md security : clarify about AI-generated reports (#26579) 2026-08-05 13:27:06 +02:00
ty.toml examples : clean-up old/obsolete stuff (#32184) 2026-11-13 10:33:17 +02:00

llama.cpp

llama.cpp

Quick start

A few options to get llama.cpp installed on your machine:

# curl
curl -LsSf https://llama.app/install.sh | sh

# powershell
irm https://llama.app/install.ps1 | iex

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-GGUF
VLM session with `llama cli` VLM session with llama cli Built-in web UI against `llama serve` running Qwen 3.6 Built-in web UI against llama serve

Description

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 Huawei Ascend NPU
CUDA Nvidia GPU
HIP AMD GPU
Hexagon 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

Contributing

  • Contributors can open PRs
  • Collaborators will be invited based on contributions
  • Maintainers can push to branches in the llama.cpp repo and merge PRs into the master branch
  • Any help with managing issues, PRs and projects is very appreciated!
  • Read the CONTRIBUTING.md for more information

Acknowledgements

  • yhirose/cpp-httplib - Single-header HTTP server, used by llama-server - MIT license
  • nothings/stb - Single-header image format decoder, used by multimodal subsystem - Public domain
  • nlohmann/json - Single-header JSON library, used by various tools/examples - MIT License
  • mackron/miniaudio - Single-header audio format decoder, used by multimodal subsystem - Public domain
  • sheredom/subprocess.h - Single-header process launching solution for C and C++ - Public domain