Coding Tools

llama.cpp b11027 adds CUDA 13 and ROCm 10 builds

llama.cpp ReleasesSeptember 18, 2026high confidence

Why it matters

→CUDA 13 support allows users with newer NVIDIA GPUs to run inference without compiling from source.
→ROCm 10.0 builds provide native AMD GPU acceleration, reducing the gap with CUDA performance.
→Pre-built binaries for multiple OS and hardware combinations simplify deployment for developers.

llama.cpp has released version b11027, adding pre-built binaries for CUDA 13 and ROCm 10.0 on Linux and Windows. The release includes CUDA 13.3 libraries for Ubuntu x64/arm64 and CUDA 13.4 DLLs for Windows, alongside existing CUDA 12 support. AMD ROCm 10.0 builds are now available for both platforms, expanding hardware compatibility. Notably, macOS KleidiAI support is disabled in this build, while openEuler ACL Graph support remains disabled. This update allows users with modern GPU drivers to run local models without manual compilation.

Read original

llama.cpp b11027 adds CUDA 13 and ROCm 10 builds

Why it matters

More from llama.cpp Releases

llama.cpp b11017 adds CUDA 13 and ROCm 10 builds

llama.cpp b11018 adds CUDA 13 and ROCm 10 builds

More in Coding Tools

Claude Code v2.1.275 patches stability and sync

Claude Code v2.1.276 fixes proxy regression

llama.cpp b11019 fixes embedded GGUF loading bugs

GitHub Copilot adds feature-level engagement metrics