# llama.cpp adds SYCL support for mixed GPU models

> The open‑source Llama.cpp library released version b11464 on Oct 7, 2026, fixing a SYCL issue that blocked running on different GPU models together.

Oossa · 2026-10-07 · https://oossa.com/en/llama-cpp-adds-sycl-support-for-mixed-gpu-models

The llama.cpp project published release b11464 on 7 Oct 2026. The update fixes a SYCL (a standard for heterogeneous computing) bug that prevented mixed GPU models from running together. The change is in the file ggml‑sycl.cpp and was co‑authored by Georgi Gerganov. The release also lists many build targets, from macOS Apple Silicon to Linux CUDA 13.4 and Android Snapdragon platforms.

## The facts

- Release tag b11464 published on 07 Oct 2026
- Fix targets SYCL FP32/FP16 builds on mixed GPU models

## Why it matters

Developers can now run Llama.cpp on systems with different GPUs without errors, expanding hardware compatibility.

## Sources & references

1. [ggml-org/llama.cpp b11464](https://github.com/ggml-org/llama.cpp/releases/tag/b11464) – llama.cpp, 2026-10-07

Last updated: 2026-10-07