# llama.cpp adds OpenCL depthwise conv2d support in v b11827

> The open‑source llama.cpp library released version b11827 with new OpenCL depthwise convolution kernels and binaries for macOS, iOS and Linux.

Oossa · 2026-10-22 · https://oossa.com/en/llama-cpp-adds-opencl-depthwise-conv2d-support-in-v-b11827

The ggml‑org team released llama.cpp version b11827. The update adds OpenCL depthwise conv2d kernels, including support for Qualcomm Adreno GPUs. Pre‑built binaries are now available for macOS Apple Silicon, macOS Intel, iOS, and several Ubuntu platforms (CPU and Vulkan). The changes were contributed by OpenCode and Qualcomm engineer Li He.

## The facts

- Release tag: b11827
- Release date: Thu Oct 22 2026

## Why it matters

Developers can now run llama.cpp models faster on devices with OpenCL GPUs, expanding low‑power hardware options.

## Sources & references

1. [ggml-org/llama.cpp b11827](https://github.com/ggml-org/llama.cpp/releases/tag/b11827) – llama.cpp, 2026-10-22

Last updated: 2026-10-22