ggml-org/llama.cppPublic

LLM inference in C/C++

AI summary: A highly optimized, plain C/C++ engine enabling rapid, local inference of large language models across diverse consumer hardware.

Star on GitHub
Stars
130.3K
+66 today
Forks
24.1K
Watchers
841
Open issues
868
Open PRs
1.7K
Contributors
~2.1K
Commits
11.3K
Branches
828

C++MITCreated Mar 10, 2023Last push todayLatest release v0.5.0+589 stars this week+3.2K this month

Quick answers

What is llama.cpp?
A highly optimized, plain C/C++ engine enabling rapid, local inference of large language models across diverse consumer hardware.
What does llama.cpp do?
llama.cpp fundamentally redefines the accessibility of Large Language Models by providing a hyper-optimized inference engine built completely in plain C/C++. The project leverages custom tensor operations and aggressive low-bit integer quantization to allow models that typically require enterprise-grade GPUs to execute with near-native performance on standard laptops and edge devices. It boasts deep, first-class hardware optimizations, natively targeting frameworks like Apple's Metal/Accelerate for Mac devices and offering robust AVX support for x86 architectures. Additionally, the engine facilitates hybrid CPU/GPU inference, ensuring expansive models can be efficiently offloaded and accelerated even on highly constrained consumer hardware.
Who is llama.cpp for?
Engineers, researchers, and hobbyists looking to run, embed, or test high-performance language models locally without relying on external cloud APIs or requiring high-end GPUs.
How do I get started with llama.cpp?
llama cli -hf ggml-org/Qwen3.5-0.8B-GGUF
How popular is llama.cpp on GitHub?
ggml-org/llama.cpp has 130,295 stars and 24,074 forks on GitHub, and gained 589 stars in the last 7 days.
What license does llama.cpp use?
ggml-org/llama.cpp is released under the MIT license.

Star history

since Jul 29, 2026
050K100KJul 2026Aug 2026Sep 2026Oct 2026
130.3K stars as of Oct 4, 2026. Measured daily since Jul 29, 2026; GitHub no longer exposes earlier star timestamps.

Contribution activity

commits per day, last 52 weeks
OctNovDecJanFebMarAprMayJunJulAugSepMonWedFri2025-10-05: 2 commits2025-10-06: 9 commits2025-10-07: 13 commits2025-10-08: 7 commits2025-10-09: 16 commits2025-10-10: 5 commits2025-10-11: 6 commits2025-10-12: 10 commits2025-10-13: 13 commits2025-10-14: 10 commits2025-10-15: 13 commits2025-10-16: 9 commits2025-10-17: 12 commits2025-10-18: 3 commits2025-10-19: 4 commits2025-10-20: 14 commits2025-10-21: 4 commits2025-10-22: 10 commits2025-10-23: 7 commits2025-10-24: 5 commits2025-10-25: 6 commits2025-10-26: 8 commits2025-10-27: 14 commits2025-10-28: 12 commits2025-10-29: 14 commits2025-10-30: 22 commits2025-10-31: 19 commits2025-11-01: 14 commits2025-11-02: 13 commits2025-11-03: 13 commits2025-11-04: 13 commits2025-11-05: 14 commits2025-11-06: 8 commits2025-11-07: 13 commits2025-11-08: 13 commits2025-11-09: 12 commits2025-11-10: 18 commits2025-11-11: 12 commits2025-11-12: 16 commits2025-11-13: 20 commits2025-11-14: 9 commits2025-11-15: 15 commits2025-11-16: 6 commits2025-11-17: 9 commits2025-11-18: 7 commits2025-11-19: 13 commits2025-11-20: 16 commits2025-11-21: 7 commits2025-11-22: 4 commits2025-11-23: 6 commits2025-11-24: 15 commits2025-11-25: 7 commits2025-11-26: 13 commits2025-11-27: 12 commits2025-11-28: 19 commits2025-11-29: 11 commits2025-11-30: 13 commits2025-12-01: 13 commits2025-12-02: 22 commits2025-12-03: 20 commits2025-12-04: 21 commits2025-12-05: 20 commits2025-12-06: 22 commits2025-12-07: 5 commits2025-12-08: 15 commits2025-12-09: 17 commits2025-12-10: 17 commits2025-12-11: 8 commits2025-12-12: 17 commits2025-12-13: 16 commits2025-12-14: 12 commits2025-12-15: 17 commits2025-12-16: 28 commits2025-12-17: 27 commits2025-12-18: 15 commits2025-12-19: 10 commits2025-12-20: 7 commits2025-12-21: 10 commits2025-12-22: 19 commits2025-12-23: 13 commits2025-12-24: 12 commits2025-12-25: 5 commits2025-12-26: 10 commits2025-12-27: 7 commits2025-12-28: 8 commits2025-12-29: 17 commits2025-12-30: 12 commits2025-12-31: 12 commits2026-01-01: 5 commits2026-01-02: 16 commits2026-01-03: 8 commits2026-01-04: 7 commits2026-01-05: 13 commits2026-01-06: 17 commits2026-01-07: 16 commits2026-01-08: 14 commits2026-01-09: 11 commits2026-01-10: 11 commits2026-01-11: 9 commits2026-01-12: 9 commits2026-01-13: 11 commits2026-01-14: 13 commits2026-01-15: 16 commits2026-01-16: 16 commits2026-01-17: 3 commits2026-01-18: 9 commits2026-01-19: 8 commits2026-01-20: 6 commits2026-01-21: 15 commits2026-01-22: 18 commits2026-01-23: 7 commits2026-01-24: 9 commits2026-01-25: 9 commits2026-01-26: 6 commits2026-01-27: 12 commits2026-01-28: 18 commits2026-01-29: 16 commits2026-01-30: 19 commits2026-01-31: 6 commits2026-02-01: 4 commits2026-02-02: 15 commits2026-02-03: 13 commits2026-02-04: 12 commits2026-02-05: 12 commits2026-02-06: 14 commits2026-02-07: 5 commits2026-02-08: 7 commits2026-02-09: 10 commits2026-02-10: 15 commits2026-02-11: 17 commits2026-02-12: 15 commits2026-02-13: 27 commits2026-02-14: 16 commits2026-02-15: 9 commits2026-02-16: 13 commits2026-02-17: 13 commits2026-02-18: 11 commits2026-02-19: 24 commits2026-02-20: 10 commits2026-02-21: 7 commits2026-02-22: 8 commits2026-02-23: 9 commits2026-02-24: 9 commits2026-02-25: 10 commits2026-02-26: 15 commits2026-02-27: 9 commits2026-02-28: 6 commits2026-03-01: 2 commits2026-03-02: 8 commits2026-03-03: 3 commits2026-03-04: 12 commits2026-03-05: 14 commits2026-03-06: 21 commits2026-03-07: 5 commits2026-03-08: 15 commits2026-03-09: 22 commits2026-03-10: 18 commits2026-03-11: 37 commits2026-03-12: 21 commits2026-03-13: 19 commits2026-03-14: 24 commits2026-03-15: 19 commits2026-03-16: 24 commits2026-03-17: 17 commits2026-03-18: 15 commits2026-03-19: 39 commits2026-03-20: 17 commits2026-03-21: 11 commits2026-03-22: 13 commits2026-03-23: 20 commits2026-03-24: 13 commits2026-03-25: 35 commits2026-03-26: 17 commits2026-03-27: 16 commits2026-03-28: 17 commits2026-03-29: 6 commits2026-03-30: 13 commits2026-03-31: 22 commits2026-04-01: 26 commits2026-04-02: 17 commits2026-04-03: 26 commits2026-04-04: 7 commits2026-04-05: 6 commits2026-04-06: 11 commits2026-04-07: 25 commits2026-04-08: 22 commits2026-04-09: 26 commits2026-04-10: 20 commits2026-04-11: 7 commits2026-04-12: 6 commits2026-04-13: 17 commits2026-04-14: 16 commits2026-04-15: 12 commits2026-04-16: 21 commits2026-04-17: 9 commits2026-04-18: 7 commits2026-04-19: 14 commits2026-04-20: 13 commits2026-04-21: 19 commits2026-04-22: 18 commits2026-04-23: 24 commits2026-04-24: 16 commits2026-04-25: 11 commits2026-04-26: 9 commits2026-04-27: 10 commits2026-04-28: 14 commits2026-04-29: 21 commits2026-04-30: 11 commits2026-05-01: 16 commits2026-05-02: 7 commits2026-05-03: 6 commits2026-05-04: 10 commits2026-05-05: 14 commits2026-05-06: 16 commits2026-05-07: 23 commits2026-05-08: 31 commits2026-05-09: 5 commits2026-05-10: 12 commits2026-05-11: 14 commits2026-05-12: 17 commits2026-05-13: 20 commits2026-05-14: 21 commits2026-05-15: 14 commits2026-05-16: 18 commits2026-05-17: 20 commits2026-05-18: 22 commits2026-05-19: 21 commits2026-05-20: 20 commits2026-05-21: 23 commits2026-05-22: 16 commits2026-05-23: 7 commits2026-05-24: 6 commits2026-05-25: 31 commits2026-05-26: 19 commits2026-05-27: 20 commits2026-05-28: 40 commits2026-05-29: 27 commits2026-05-30: 14 commits2026-05-31: 7 commits2026-06-01: 25 commits2026-06-02: 23 commits2026-06-03: 13 commits2026-06-04: 26 commits2026-06-05: 20 commits2026-06-06: 9 commits2026-06-07: 10 commits2026-06-08: 16 commits2026-06-09: 24 commits2026-06-10: 6 commits2026-06-11: 12 commits2026-06-12: 18 commits2026-06-13: 13 commits2026-06-14: 12 commits2026-06-15: 23 commits2026-06-16: 14 commits2026-06-17: 24 commits2026-06-18: 21 commits2026-06-19: 21 commits2026-06-20: 14 commits2026-06-21: 12 commits2026-06-22: 10 commits2026-06-23: 12 commits2026-06-24: 10 commits2026-06-25: 18 commits2026-06-26: 28 commits2026-06-27: 8 commits2026-06-28: 11 commits2026-06-29: 10 commits2026-06-30: 11 commits2026-07-01: 8 commits2026-07-02: 1 commit2026-07-03: 13 commits2026-07-04: 3 commits2026-07-05: 8 commits2026-07-06: 16 commits2026-07-07: 25 commits2026-07-08: 26 commits2026-07-09: 22 commits2026-07-10: 28 commits2026-07-11: 10 commits2026-07-12: 4 commits2026-07-13: 18 commits2026-07-14: 26 commits2026-07-15: 25 commits2026-07-16: 21 commits2026-07-17: 12 commits2026-07-18: 3 commits2026-07-19: 2 commits2026-07-20: 7 commits2026-07-21: 10 commits2026-07-22: 20 commits2026-07-23: 7 commits2026-07-24: 15 commits2026-07-25: 10 commits2026-07-26: 14 commits2026-07-27: 20 commits2026-07-28: 29 commits2026-07-29: 10 commits2026-07-30: 19 commits2026-07-31: 23 commits2026-08-01: 6 commits2026-08-02: 15 commits2026-08-03: 31 commits2026-08-04: 23 commits2026-08-05: 15 commits2026-08-06: 13 commits2026-08-07: 26 commits2026-08-08: 5 commits2026-08-09: 7 commits2026-08-10: 33 commits2026-08-11: 27 commits2026-08-12: 27 commits2026-08-13: 25 commits2026-08-14: 24 commits2026-08-15: 24 commits2026-08-16: 7 commits2026-08-17: 32 commits2026-08-18: 20 commits2026-08-19: 28 commits2026-08-20: 26 commits2026-08-21: 28 commits2026-08-22: 19 commits2026-08-23: 20 commits2026-08-24: 13 commits2026-08-25: 18 commits2026-08-26: 12 commits2026-08-27: 25 commits2026-08-28: 26 commits2026-08-29: 11 commits2026-08-30: 21 commits2026-08-31: 26 commits2026-09-01: 20 commits2026-09-02: 20 commits2026-09-03: 27 commits2026-09-04: 21 commits2026-09-05: 6 commits2026-09-06: 13 commits2026-09-07: 29 commits2026-09-08: 13 commits2026-09-09: 21 commits2026-09-10: 20 commits2026-09-11: 23 commits2026-09-12: 16 commits2026-09-13: 16 commits2026-09-14: 23 commits2026-09-15: 19 commits2026-09-16: 29 commits2026-09-17: 17 commits2026-09-18: 20 commits2026-09-19: 16 commits2026-09-20: 7 commits2026-09-21: 34 commits2026-09-22: 21 commits2026-09-23: 33 commits2026-09-24: 30 commits2026-09-25: 23 commits2026-09-26: 14 commits2026-09-27: 17 commits2026-09-28: 15 commits2026-09-29: 15 commits2026-09-30: 29 commits2026-10-01: 38 commits2026-10-02: 0 commits2026-10-03: 0 commits
5,510 commits in the last yearLessMore

Signals and awards

derived from tracked data

  • Landmark project

    130,295 stars

  • Very active

    5,510 commits in 52 weeks

  • Community-driven

    ~2,074 contributors

  • Permissive license

    MIT

  • Continuous integration

    Automated checks passing

  • Top 10% tracked

    Rank 50 of 1135

What llama.cpp does

llama.cpp fundamentally redefines the accessibility of Large Language Models by providing a hyper-optimized inference engine built completely in plain C/C++. The project leverages custom tensor operations and aggressive low-bit integer quantization to allow models that typically require enterprise-grade GPUs to execute with near-native performance on standard laptops and edge devices. It boasts deep, first-class hardware optimizations, natively targeting frameworks like Apple's Metal/Accelerate for Mac devices and offering robust AVX support for x86 architectures. Additionally, the engine facilitates hybrid CPU/GPU inference, ensuring expansive models can be efficiently offloaded and accelerated even on highly constrained consumer hardware.

Engineers, researchers, and hobbyists looking to run, embed, or test high-performance language models locally without relying on external cloud APIs or requiring high-end GPUs.

  • Extreme Portability: Engineered completely in C/C++ without external dependencies, allowing trivial cross-compilation for nearly any platform.
  • Aggressive Quantization: Shrinks model weights using advanced integer quantization from 1.5-bit to 8-bit to slash memory footprints rapidly.
  • Apple Silicon Native: Features premier integration with ARM NEON and Apple's Accelerate frameworks to maximize M-series chip performance natively.
  • Hybrid Inference Offloading: Dynamically partitions models across both the CPU and available discrete GPUs to overcome strict VRAM limitations.
  • Extensive Hardware Backends: Out-of-the-box support for an array of backends including Vulkan, SYCL, OpenCL, CUDA, and HIP architectures.
  • Embedded HTTP Server: Ships with a lightweight, built-in REST API server enabling instant OpenAI-compatible web endpoint deployments.

Where teams use it

Local Chatbots

Developers use the engine as the highly efficient backend for offline, privacy-focused AI assistants that run fluidly on standard laptops.

Edge Computing

Engineers deploy heavily quantized language models onto constrained edge devices or IoT hardware to guarantee internet-free operations.

App Integration

Software teams utilize the engine's provided multi-language bindings and API server to easily embed robust LLM logic directly into existing native applications.

Cross-Platform AI Agents

Researchers leverage the single-binary architecture to distribute powerful AI agents seamlessly across wildly different hardware landscapes without massive dependency chains.

Getting started: llama cli -hf ggml-org/Qwen3.5-0.8B-GGUF

README

master branch

llama.cpp

llama

Quick start

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

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 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
View on GitHub

Recent activity

commits and pull requests

Recent open issues

view all

Discussions

all 3.7K

Releases and announcements

1,000 total
  1. b11326b11326Oct 1, 2026pre-release182 downloads

    <details open> meta: clear inactive AllReduce shards with FILL, not SCALE (#29793) </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/51949614> **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11326/llama-b11326-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11326/llama-b11326-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11326/llama-b11326-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11326/llama-b11326-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11326/llama-b11326-bin-ubuntu-arm64.tar.gz) - [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11326/llama-b11326-bin-ubuntu-s390x.tar.gz) - [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/b11326/llama-b11326-bin-ubuntu-vulk

  2. b11325b11325Oct 1, 2026pre-release559 downloads

    <details open> jinja : skip copying loop scope unless a loop filter needs it (#29776) </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/51940374> **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11325/llama-b11325-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11325/llama-b11325-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11325/llama-b11325-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11325/llama-b11325-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11325/llama-b11325-bin-ubuntu-arm64.tar.gz) - [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11325/llama-b11325-bin-ubuntu-s390x.tar.gz) - [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/b11325/llama-b11325-bin-ubuntu-v

  3. b11324b11324Oct 1, 2026pre-release664 downloads

    <details open> llama-mmap : avoid a second full-size copy of each tensor with direct-io (#29749) Assisted-by: Claude Co-authored-by: Pranesh Gonegandla <pgonegandla@nvidia.com> </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/51929570> **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11324/llama-b11324-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11324/llama-b11324-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11324/llama-b11324-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11324/llama-b11324-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11324/llama-b11324-bin-ubuntu-arm64.tar.gz) - [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11324/llama-b11324-bin-ubuntu-s390x.tar.gz) - [Ubuntu x64 (Vulk

  4. b11323b11323Oct 1, 2026pre-release876 downloads

    <details open> HIP: avoid treating CDNA as dgx spark for gqa_ratio 20 in fattn_mma dqk 576 (#29572) </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/51919562> **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11323/llama-b11323-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11323/llama-b11323-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11323/llama-b11323-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11323/llama-b11323-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11323/llama-b11323-bin-ubuntu-arm64.tar.gz) - [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11323/llama-b11323-bin-ubuntu-s390x.tar.gz) - [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/b11323/llama-b1132

  5. b11322b11322Oct 1, 2026pre-release1.2K downloads

    <details open> hex-workqueue: fix race condition in seqn getting out of sync with idx_read/write (#29785) </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/51906249> **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11322/llama-b11322-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11322/llama-b11322-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11322/llama-b11322-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11322/llama-b11322-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11322/llama-b11322-bin-ubuntu-arm64.tar.gz) - [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11322/llama-b11322-bin-ubuntu-s390x.tar.gz) - [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/b11322/llama

Commits per week

last 52 weeks
1670Week of 2025-10-05: 58 commitsWeek of 2025-10-12: 70 commitsWeek of 2025-10-19: 50 commitsWeek of 2025-10-26: 103 commitsWeek of 2025-11-02: 87 commitsWeek of 2025-11-09: 102 commitsWeek of 2025-11-16: 62 commitsWeek of 2025-11-23: 83 commitsWeek of 2025-11-30: 131 commitsWeek of 2025-12-07: 95 commitsWeek of 2025-12-14: 116 commitsWeek of 2025-12-21: 76 commitsWeek of 2025-12-28: 78 commitsWeek of 2026-01-04: 89 commitsWeek of 2026-01-11: 77 commitsWeek of 2026-01-18: 72 commitsWeek of 2026-01-25: 86 commitsWeek of 2026-02-01: 75 commitsWeek of 2026-02-08: 107 commitsWeek of 2026-02-15: 87 commitsWeek of 2026-02-22: 66 commitsWeek of 2026-03-01: 65 commitsWeek of 2026-03-08: 156 commitsWeek of 2026-03-15: 142 commitsWeek of 2026-03-22: 131 commitsWeek of 2026-03-29: 117 commitsWeek of 2026-04-05: 117 commitsWeek of 2026-04-12: 88 commitsWeek of 2026-04-19: 115 commitsWeek of 2026-04-26: 88 commitsWeek of 2026-05-03: 105 commitsWeek of 2026-05-10: 116 commitsWeek of 2026-05-17: 129 commitsWeek of 2026-05-24: 157 commitsWeek of 2026-05-31: 123 commitsWeek of 2026-06-07: 99 commitsWeek of 2026-06-14: 129 commitsWeek of 2026-06-21: 98 commitsWeek of 2026-06-28: 57 commitsWeek of 2026-07-05: 135 commitsWeek of 2026-07-12: 109 commitsWeek of 2026-07-19: 71 commitsWeek of 2026-07-26: 121 commitsWeek of 2026-08-02: 128 commitsWeek of 2026-08-09: 167 commitsWeek of 2026-08-16: 160 commitsWeek of 2026-08-23: 125 commitsWeek of 2026-08-30: 141 commitsWeek of 2026-09-06: 135 commitsWeek of 2026-09-13: 140 commitsWeek of 2026-09-20: 162 commitsWeek of 2026-09-27: 114 commitsOct 5, 2025Sep 27, 2026
5.5K commits in the last 52 weeks.

When work happens

weekday and hour
SunMonTueWedThuFriSat036912151821Sun 0:00 — 41 commitsSun 1:00 — 32 commitsSun 2:00 — 21 commitsSun 3:00 — 37 commitsSun 4:00 — 18 commitsSun 5:00 — 14 commitsSun 6:00 — 21 commitsSun 7:00 — 29 commitsSun 8:00 — 49 commitsSun 9:00 — 67 commitsSun 10:00 — 53 commitsSun 11:00 — 50 commitsSun 12:00 — 56 commitsSun 13:00 — 52 commitsSun 14:00 — 67 commitsSun 15:00 — 66 commitsSun 16:00 — 76 commitsSun 17:00 — 56 commitsSun 18:00 — 82 commitsSun 19:00 — 75 commitsSun 20:00 — 56 commitsSun 21:00 — 49 commitsSun 22:00 — 63 commitsSun 23:00 — 52 commitsMon 0:00 — 37 commitsMon 1:00 — 38 commitsMon 2:00 — 30 commitsMon 3:00 — 31 commitsMon 4:00 — 18 commitsMon 5:00 — 24 commitsMon 6:00 — 29 commitsMon 7:00 — 39 commitsMon 8:00 — 72 commitsMon 9:00 — 92 commitsMon 10:00 — 114 commitsMon 11:00 — 90 commitsMon 12:00 — 90 commitsMon 13:00 — 121 commitsMon 14:00 — 136 commitsMon 15:00 — 110 commitsMon 16:00 — 111 commitsMon 17:00 — 97 commitsMon 18:00 — 79 commitsMon 19:00 — 93 commitsMon 20:00 — 67 commitsMon 21:00 — 70 commitsMon 22:00 — 89 commitsMon 23:00 — 76 commitsTue 0:00 — 47 commitsTue 1:00 — 47 commitsTue 2:00 — 23 commitsTue 3:00 — 35 commitsTue 4:00 — 30 commitsTue 5:00 — 27 commitsTue 6:00 — 39 commitsTue 7:00 — 41 commitsTue 8:00 — 73 commitsTue 9:00 — 104 commitsTue 10:00 — 104 commitsTue 11:00 — 107 commitsTue 12:00 — 87 commitsTue 13:00 — 98 commitsTue 14:00 — 95 commitsTue 15:00 — 119 commitsTue 16:00 — 95 commitsTue 17:00 — 96 commitsTue 18:00 — 89 commitsTue 19:00 — 88 commitsTue 20:00 — 68 commitsTue 21:00 — 74 commitsTue 22:00 — 64 commitsTue 23:00 — 65 commitsWed 0:00 — 52 commitsWed 1:00 — 60 commitsWed 2:00 — 42 commitsWed 3:00 — 23 commitsWed 4:00 — 24 commitsWed 5:00 — 22 commitsWed 6:00 — 32 commitsWed 7:00 — 50 commitsWed 8:00 — 69 commitsWed 9:00 — 91 commitsWed 10:00 — 99 commitsWed 11:00 — 92 commitsWed 12:00 — 106 commitsWed 13:00 — 111 commitsWed 14:00 — 137 commitsWed 15:00 — 117 commitsWed 16:00 — 112 commitsWed 17:00 — 88 commitsWed 18:00 — 99 commitsWed 19:00 — 93 commitsWed 20:00 — 86 commitsWed 21:00 — 58 commitsWed 22:00 — 83 commitsWed 23:00 — 81 commitsThu 0:00 — 34 commitsThu 1:00 — 58 commitsThu 2:00 — 46 commitsThu 3:00 — 36 commitsThu 4:00 — 21 commitsThu 5:00 — 25 commitsThu 6:00 — 29 commitsThu 7:00 — 41 commitsThu 8:00 — 87 commitsThu 9:00 — 112 commitsThu 10:00 — 118 commitsThu 11:00 — 109 commitsThu 12:00 — 102 commitsThu 13:00 — 117 commitsThu 14:00 — 105 commitsThu 15:00 — 125 commitsThu 16:00 — 95 commitsThu 17:00 — 95 commitsThu 18:00 — 85 commitsThu 19:00 — 107 commitsThu 20:00 — 72 commitsThu 21:00 — 85 commitsThu 22:00 — 67 commitsThu 23:00 — 74 commitsFri 0:00 — 53 commitsFri 1:00 — 46 commitsFri 2:00 — 40 commitsFri 3:00 — 25 commitsFri 4:00 — 26 commitsFri 5:00 — 31 commitsFri 6:00 — 44 commitsFri 7:00 — 35 commitsFri 8:00 — 81 commitsFri 9:00 — 111 commitsFri 10:00 — 108 commitsFri 11:00 — 119 commitsFri 12:00 — 110 commitsFri 13:00 — 110 commitsFri 14:00 — 117 commitsFri 15:00 — 124 commitsFri 16:00 — 90 commitsFri 17:00 — 96 commitsFri 18:00 — 76 commitsFri 19:00 — 70 commitsFri 20:00 — 86 commitsFri 21:00 — 74 commitsFri 22:00 — 46 commitsFri 23:00 — 49 commitsSat 0:00 — 40 commitsSat 1:00 — 42 commitsSat 2:00 — 39 commitsSat 3:00 — 30 commitsSat 4:00 — 27 commitsSat 5:00 — 23 commitsSat 6:00 — 21 commitsSat 7:00 — 32 commitsSat 8:00 — 47 commitsSat 9:00 — 62 commitsSat 10:00 — 60 commitsSat 11:00 — 66 commitsSat 12:00 — 70 commitsSat 13:00 — 62 commitsSat 14:00 — 59 commitsSat 15:00 — 81 commitsSat 16:00 — 90 commitsSat 17:00 — 78 commitsSat 18:00 — 68 commitsSat 19:00 — 47 commitsSat 20:00 — 53 commitsSat 21:00 — 66 commitsSat 22:00 — 53 commitsSat 23:00 — 38 commits
Commit volume by weekday and hour (UTC). Larger dots mean more commits.

Trending appearances

DateListRankStars gained
Jul 13, 2026daily#24+3
May 19, 2026daily#24+30

Related repositories