# llama.cpp b12050 tunes Vulkan for newer AMD GPUs

> The release adjusts Vulkan split-K selection for small workloads on AMD’s RDNA 3 and RDNA 4 GPUs. It also offers downloadable builds for several platforms.

Oossa · 2026-11-02 · https://oossa.com/en/llama-cpp-b12050-tunes-vulkan-for-newer-amd-gpus

The ggml-org team has released llama.cpp b12050, with a Vulkan change aimed at improving how the software handles small workloads on AMD RDNA 3 and RDNA 4 graphics cards. The release notes do not give benchmark results or quantify any speed-up.

Downloadable builds include packages for macOS, iOS and several Linux CPU platforms.

## The facts

- The change is listed as pull request #30146.
- Linux downloads include Ubuntu x64, arm64 and s390x CPU builds.

## Why it matters

People running llama.cpp on RDNA 3 or RDNA 4 GPUs may get better handling of small workloads, but the release notes do not say how much performance changes in practice.

## Sources & references

1. [ggml-org/llama.cpp b12050](https://github.com/ggml-org/llama.cpp/releases/tag/b12050) – llama.cpp, 2026-11-02

Last updated: 2026-11-02