Coding Tools

llama.cpp b11042 adds ROCm 10 and CUDA 13 builds

llama.cpp ReleasesSeptember 19, 2026high confidence

Why it matters

→ROCm 10 support brings AMD GPU users to parity with CUDA in official binaries.
→CUDA 13 availability future-proofs local inference against upcoming NVIDIA driver updates.
→OpenCL kernel additions improve performance on non-NVIDIA accelerators like Intel Arc.

llama.cpp has released version b11042, adding official build support for ROCm 10.0 and CUDA 13 (up to 13.4) on Linux and Windows platforms. The update includes new OpenCL binary kernels for Q6_K quantization and fixes layout compatibility issues. macOS KleidiAI builds are currently disabled in this release. This version ensures compatibility with the latest driver stacks from both AMD and NVIDIA.

Read original

llama.cpp b11042 adds ROCm 10 and CUDA 13 builds

Why it matters

More from llama.cpp Releases

llama.cpp b11036 adds CUDA 13 and ROCm 10.0

llama.cpp b11037 fixes GPU inference bugs

More in Coding Tools

Claude Code v2.1.277 fixes critical bugs and adds AGENTS.md

Claude Code v2.1.278 optimizes auto-mode billing

llama.cpp b11039 fixes SWA pattern roundtrip

GitHub Copilot Code Review GA with Auto-Resolution