Note
This is p100-series: llama.cpp tuned for NVIDIA Tesla P100 (Pascal, sm_60) GPUs.
It is upstream llama.cpp plus a linear stack of Pascal patches: faster decode matvecs, two-GPU tensor split with an in-kernel
P2P AllReduce, an MoE expert cache for models larger than VRAM, and Qwen 3.8 Flash-Next support. The stack is replayed onto new
upstream commits and gated on real P100s (correctness, KL divergence, speed, serving) before each p100-series-vX.Y tag.
Other GPUs keep upstream's code paths.
Current release: p100-series-v5.40 (on upstream 9c9074e14, 2026-11-06).
- What's in it and how to build: P100-SERIES.md
- Measured gains, step by step vs stock llama.cpp (decode up to 1.7x, plus better accuracy): docs/p100/PERFORMANCE.md
- Much of this builds on other people's work: CREDITS.md
- Contributing (AI-written PRs welcome here): CONTRIBUTING.md
Who wrote the 225 code patches in p100-series-v5.40 (git authorship; carried commits link their originals):
| Author | Patches | What |
|---|---|---|
| wispoffates (maintainer, with Claude Code) | 177 | Pascal kernels, two-GPU tensor-split engine, expert cache, Flash-Next, checkpoint staging; 5 of them also credit qr0n, buun or Keoian as co-author / source |
| Kaden (Kmic-68) | 33 | the original sm_60 matvec rebuild, fp16 Q6_K verify matvecs, cuBLAS ALGO6, sampler prefilter, fp32 fold of the tile-FA accumulator |
| drrros | 5 | pipeline parallelism with MoE experts in host memory (upstream PR #29963) |
| Keoian (via Kmic-68) | 4 | faster repetition penalties, LoRA under tensor split, cheaper chat-turn checkpoints |
| Kane | 3 | gated-delta-net decode chain fused into 4 kernels, IQ matvec tuning for sm_60, MoE MMQ tile sized per expert |
| lich-77 | 2 | the GPU-resident MoE expert LRU cache our expert cache builds on |
| wrj97 | 1 | draft KV kept out of context checkpoints (upstream PR #29509) |
Everything below this note is upstream's README.
LLM inference in C/C++
ggml / ops / maintainer PRs / dev stats / lib llama API / llama-server REST API
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- 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-GGUF
VLM session with llama cli
|
Built-in web UI against llama serve
|
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.
| 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 |
- How to build
- Running on Docker
- Build on Android
- Multi-GPU usage
- Performance troubleshooting
- GGML tips & tricks
- XCFramework
- Completions
- Models
- Release process
- Contributors can open PRs
- Collaborators will be invited based on contributions
- Maintainers can push to branches in the
llama.cpprepo and merge PRs into themasterbranch - Any help with managing issues, PRs and projects is very appreciated!
- Read the CONTRIBUTING.md for more information
- 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