Skip to content

llama.cpp b9637 - GPU Accelerated Binaries

Latest

Choose a tag to compare

@solarpunked solarpunked released this 13 Jul 03:15

Static llama-server binaries compiled from llama.cpp b9637

Built for:

  • Linux x86_64 (Vulkan GPU + CPU fallback)
  • macOS Apple Silicon (Metal GPU)
  • macOS Intel (CPU only)
  • Windows x64 (Vulkan GPU + CPU fallback)

GPU Support:

  • Linux & Windows: Vulkan — works with NVIDIA, AMD, and Intel GPUs
  • macOS ARM: Metal — native Apple Silicon acceleration
  • macOS Intel: CPU only (no GPU acceleration)
  • All platforms fall back to CPU if no compatible GPU is found

Optional CUDA engine (llama-server-cuda-*.zip):

  • Faster generation on NVIDIA GPUs, CUDA runtime included (Linux: CUDA 12.2, drivers R525+; Windows: CUDA 12.9, drivers R560+)
  • Downloaded on demand from Settings — never bundled in installers

Configuration:

  • WebUI excluded (-DLLAMA_SERVER_WEBUI=OFF)
  • No CURL dependency (-DLLAMA_CURL=OFF)

Built on: $(date -u)