Skip to content

v0.3.20 — Blackwell (sm_100 + sm_120), CUDA 13.0, Python 3.10–3.13

Latest
Latest

Choose a tag to compare

View all tags
@dougeeai dougeeai released this 18 Apr 19:43
v0.3.20-cuda13.0-sm100-sm120
806790a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Pre-built llama-cpp-python wheel for Windows with CUDA 13.0 support — all Blackwell GPUs in a single wheel (datacenter sm_100 + consumer/workstation sm_120).

Skip the build process entirely. This wheel is compiled and ready to install.

Requirements

  • OS: Windows 10/11 64-bit
  • Python: 3.10, 3.11, 3.12, or 3.13 (single wheel works across all — py3-none ABI, loaded via ctypes)
  • CUDA: CUDA 13.0 Toolkit installed (required for cublas64_13.dll — the NVIDIA driver alone does not ship this DLL)
  • Driver: NVIDIA Driver 580 or higher
  • GPU: NVIDIA RTX 5090, 5080, 5070 Ti, 5070, 5060 Ti, 5060, 5050, RTX 5090 Laptop, RTX 5080 Laptop, RTX 5070 Ti Laptop, RTX 5070 Laptop, RTX 5060 Laptop, RTX 5050 Laptop, RTX PRO 6000 Blackwell Workstation Edition, RTX PRO 6000 Blackwell Max-Q, RTX PRO 6000 Blackwell Server Edition, RTX PRO 5000 Blackwell, RTX PRO 4500 Blackwell, RTX PRO 4000 Blackwell, RTX PRO 4000 SFF Blackwell, RTX PRO 2000 Blackwell, RTX PRO 5000 Blackwell Laptop, RTX PRO 4000 Blackwell Laptop, RTX PRO 3000 Blackwell Laptop, RTX PRO 2000 Blackwell Laptop, RTX PRO 1000 Blackwell Laptop, RTX PRO 500 Blackwell Laptop, B100, B200, B300 (Blackwell Ultra), GB200, GB300
  • Architecture: Blackwell — sm_100 (datacenter) + sm_120 (consumer/workstation) compiled into one wheel
  • VRAM: 8GB+ recommended

Installation

pip install llama_cpp_python-0.3.20+cuda13.0.sm100.sm120.blackwell-py3-none-win_amd64.whl

What This Solves

  • No Visual Studio required
  • No compilation errors
  • No "No CUDA toolset found" issues
  • No need to pick a wheel per Python version (3.10 → 3.13 all use this one)
  • No need to pick a wheel per Blackwell product line (consumer, workstation, datacenter all supported)
  • Works immediately with GGUF models

Keywords

llama-cpp-python, 0.3.20, CUDA 13.0, Python 3.10, 3.11, 3.12, 3.13, Windows, RTX 5090, RTX 5080, RTX 5070 Ti, RTX 5070, RTX 5060, RTX 5050, RTX PRO 6000 Blackwell, RTX PRO 5000 Blackwell, RTX PRO 4500 Blackwell, RTX PRO 4000 Blackwell, RTX PRO 2000 Blackwell, B100, B200, B300, Blackwell Ultra, GB200, GB300, Blackwell, sm100, sm120, consumer Blackwell, datacenter Blackwell, workstation Blackwell, GGUF, llama.cpp, no compilation, prebuilt wheel, Windows 11, Windows 10

Assets 3
Loading